Technology Archives - International Finance

Anthropic staff warn of doom while company hunts spies, hackers and bioweaponeers

International Finance Business Desk — Thu, 17 Sep 2026 00:00:18 +0000

When an artificial intelligence (AI) company publishes a report showing that its own product has been used to help design missiles, run spy operations and probe the edges of bioweapons research, it invites an obvious question.

Is this a company that has lost control of its technology, or one that is proving it can police it? In September 2026 Anthropic, the maker of the Claude family of models, gave the world a fresh reason to ask.

Two stories, one company
Within days of each other, two very different narratives about Anthropic collided in public view. On one side, a former researcher walked out of the building warning that the industry could get everyone killed. On the other, the company released its most detailed account yet of how it had caught and shut down real-world abuse of Claude.

The timing was striking. On September 8 2026, Jacob Coxon, a 27-year-old researcher who said he had spent three years doing pretraining research at both OpenAI and Anthropic, announced his resignation on X.

“I resigned from Anthropic today. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives,” he wrote.

His seven-part thread drew nearly 76 million views overnight and, by later counts, more than 170 million.

Speaking to CNN’s Anderson Cooper days afterwards, he warned that advanced systems could “cause extreme havoc, for example, hacking critical infrastructure, building extinction-level bioweapons, and said many people building the technology earnestly believe that it could kill us all by the end of the decade.”

Then, on September 10, Anthropic published its fourth threat intelligence report since March 2025, titled “Detecting and countering misuse of AI.” It read like a charge sheet.

What Anthropic says it caught
The report covered activity disrupted between December 2025 and August 2026 across seven areas of harm, including cyber operations, influence operations, surveillance, scams, biological misuse, conventional weapons development and illicit model distillation.

The cases are eye-catching. Anthropic said a China-based actor used Claude to build electronic-warfare and air-defence suppression software, and at one point altered a simulation to include 12 targets in Taiwan, among them early-warning radar, Patriot and Tien Kung missile batteries, air bases and a command bunker.

Another China-based actor, assessed to be linked to a defence manufacturer, used Claude to help write a technical proposal of more than 200 pages for an anti-torpedo system for the Chinese navy.

A cell in northern Yemen used Claude to help develop a guided rocket and a planned ballistic missile with a range of more than 2,000km, although Anthropic said there was no evidence a working weapon was fielded.

On cyber operations, Anthropic described a Russia-linked espionage group, tracked internally as GTG-20006, whose tradecraft matched the actor known as Midnight Blizzard, previously tied by the US government to Russia’s SVR foreign intelligence service.

The group allegedly ran phishing, hotel Wi-Fi hijacking and WhatsApp-takeover operations against Ukrainian and European government, military and diplomatic targets, and built an AI-driven system that automatically rewrote its malware whenever security tools detected it.

On distillation, Anthropic said it disrupted attacks from seven China-based labs. It named Alibaba, Moonshot, DeepSeek and Xiaomi among them.

It said operators linked to Alibaba ran the largest illicit distillation effort, allegedly aimed at extracting Claude’s capabilities to improve the Qwen models, and that it observed more than 151 million exchanges attributed to Alibaba between May and July 2026, peaking at nearly 3 million a day from more than 3,500 accounts it described as fraudulent.

The biological cases were the ones the company flagged as most serious. Anthropic said it blocked five separate attempts by researchers to use Claude in ways that could support biological weapons development, including a request to help write a grant application for gain-of-function research on the chikungunya virus, and work on highly pathogenic avian influenza focused on adaptation to mammals.

It said it could not always establish intent, but blocked the activity because the potential consequences were too serious to ignore.

Jacob Klein, Anthropic’s head of threat intelligence, told Reuters that model improvements had raised the stakes.

“A year ago, let’s say you wanted to optimise a drone or optimise the software on a missile, the models just wouldn’t be as good at that task as they are now,” he said.

In every case, Anthropic said, it banned the accounts, tightened its safeguards and shared intelligence with authorities and industry partners.

Fear inside the house
Coxon is not a lone voice. In February 2026, Mrinank Sharma, a member of Anthropic’s technical staff since 2023, resigned in an open letter that declared “the world is in peril.”

Two current employees publicly agreed with Coxon’s warning. Most strikingly, Evan Hubinger, who leads Anthropic’s alignment science work, backed the substance of Coxon’s claims and put his own estimate of AI causing human extinction at above 10% within the next decade, while stressing that today’s deployed models present comparatively low risk.

The company’s leaders have not been quiet either. Jack Clark, a co-founder, published an essay in October 2025, “Technological Optimism and Appropriate Fear,” describing powerful AI as a “creature” that its makers do not fully understand.

Chief executive Dario Amodei, who has estimated the chance of something going “quite catastrophically wrong on the scale of human civilisation” at between 10% and 25%, published a lengthy essay in January 2026 warning of civilisation-level risks.

And in July 2026 more than 1,100 employees across the leading AI firms signed the “Pacing the Frontier” letter urging governments to prepare to slow AI development.

The count stood at 1,171 at a first tally and passed 1,260 within two days.

Its signatories included Amodei and fellow Anthropic co-founders Jared Kaplan and Jack Clark, plus OpenAI chief scientist Jakub Pachocki, and both OpenAI and Anthropic endorsed it at the company level on July 29.

There is an obvious tension here. The same people warning that the technology could end humanity are the ones building and selling it, arguing that if they do not, less careful rivals will.

How serious is the threat, really?
Not everyone is convinced the danger is as sharp as Anthropic’s reports suggest. When the company disclosed an alleged Chinese AI-orchestrated hacking campaign in November 2025, several security researchers pushed back.

Dan Tentler of Phobos Group questioned why models supposedly did the attackers’ bidding when ordinary users hit refusals.

Bob Rudis of GreyNoise Intelligence said the disclosure did not “expand the threat model in a meaningful way” and mostly repackaged known trends. Critics also note the marketing incentive, since Anthropic sells Claude as a cyber-defence tool.

Yet independent evidence points the same way as Anthropic’s warnings, if more cautiously.

The UK’s AI Security Institute, in its first Frontier AI Trends Report on December 18 2025, found that “AI models can now complete apprentice-level tasks 50% of the time on average, compared to just over 10% of the time in early 2024,” and said frontier models now regularly exceed PhD-level baselines on biology and chemistry knowledge tests.

The same institute found in July 2026 that every frontier model it tested, including systems from OpenAI and Anthropic, tried to cheat on cyber evaluations without being told to.

The picture that emerges is of genuine, fast-rising capability paired with real uncertainty about control, which is closer to Anthropic’s framing than to the sceptics’ dismissal.

What governments are doing
The revelations land in a world still assembling its rulebook. In the European Union (EU), the AI Act’s obligations for general-purpose AI, in force since August 2025, become enforceable with the threat of fines from August 2 2026, although the bloc pushed some high-risk requirements back to 2027 and 2028.

Anthropic is a signatory to the EU’s voluntary code of practice.

The United States has moved the other way. On December 11 2025, President Trump signed Executive Order 14365, seeking to curb state AI laws through a Justice Department litigation task force and the threat of withheld funding, in the name of “AI dominance.”

State legislatures pressed ahead anyway. By the New York University Center on Technology Policy’s count, states had enacted 109 AI laws across 29 states by July 1 2026.

Elsewhere, the United Kingdom’s AI Security Institute continues its pre-deployment testing. China has tightened its regime, with mandatory labelling of AI-generated content from September 2025 and AI provisions folded into its amended Cybersecurity Law from January 2026.

India hosted the “AI Impact Summit” in New Delhi from February 16 to 20 2026, the first such gathering in the Global South, which adopted a Leaders’ Declaration.

The United Nations, meanwhile, established an “Independent International Scientific Panel” of 40 experts, appointed on February 12 2026 by a recorded vote of 117 to 2 and co-chaired by Yoshua Bengio and Maria Ressa, along with a “Global Dialogue on AI Governance” whose first meeting took place in Geneva in 2026.

Will anything change?
Anthropic’s report is, in effect, an argument for exactly the kind of oversight its own staff are demanding. By showing that would-be bioweaponeers, spies and hackers are already knocking, the company hands regulators concrete evidence rather than speculation.

Whether that shifts policy is another matter. The US is actively resisting binding rules, the EU is softening timelines, and the international bodies remain talking shops without enforcement power.

The uncomfortable takeaway is that the safeguards catching today’s abuse are being built and operated by the same firms racing to make the technology more powerful. For businesses, the practical lesson is not to wait for governments to settle the argument.

Treat AI credentials and agent integrations as production secrets, since Anthropic found stolen keys were a prime target.

Watch the enforcement dates that carry real teeth, above all the EU’s August 2026 deadline.

And read the frontier labs’ own threat reports closely, because for now they are the clearest public window into how this technology is actually being misused.

The post Anthropic staff warn of doom while company hunts spies, hackers and bioweaponeers appeared first on International Finance.

Altman, Amodei and Musk unite on AI safety as OpenAI defers IPO

International Finance Business Desk — Tue, 15 Sep 2026 00:00:41 +0000

Citing safety concerns over artificial intelligence, OpenAI CEO Sam Altman has ruled out an IPO for the company in 2026. Altman believes even a 10% risk of AI causing human extinction by the end of the decade would be “unacceptable.”

“I actually think that, given everything happening with safety, right now would be an ill-advised moment to go public, and we don’t feel pressure on that,” Altman told business magazine Fortune on Saturday (September 11).

While stating that he did not know how an estimate about the world facing a 10% chance of AI-driven extinction could be made, Altman still said that the risk was serious enough to make AI companies and governments act seriously on the policy front.

“Regardless of whether it’s 10, eight, or six, the key takeaway is that we all bear a significant responsibility and must not allow egos, profit motives, or any other factors to interfere. We must act to avoid those risks, and I think we can,” the OpenAI boss said.

However, unlike Anthropic, which has decided to proceed with its IPO, OpenAI has not taken many steps in that direction. In June, the New York Times (NYT) reported that the San Francisco-based venture was considering whether to ‌hold ⁠off the market listing until 2027 on a potentially trillion-dollar IPO.

Altman told Fortune, “I would say we are not aiming for 2026. We have a lot of tasks to complete, such as addressing the current requirements for safety and alignment and determining how the industry and governments can collaborate effectively.”

He ⁠also suggested that OpenAI and other leading AI firms were closing in on unveiling a pact to slow AI development and work together to tackle safety risks.

Altman also found support from Anthropic CEO Dario Amodei, with the latter urging AI companies to take a more deliberate approach to development.

“We must slow the pace at which we improve ⁠the capabilities of AI models,” Amodei wrote in an essay on X (formerly Twitter).

Agreeing with his rival CEO, Altman wrote on the popular micro-blogging platform, “I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we’ve had at OpenAI in recent weeks.”

Amodei has pitched a three-step plan, in which embedded independent evaluators will have employee-like access to verify safety practices within the AI firms, apart from paving the way for coordination among frontier AI firms to set safety standards and limit unchecked AI development, and last but not least, international cooperation to manage AI risks.

Both XAI boss Elon Musk and Altman agreed with Amodei.

“Committing to having independent evaluators with employee-like access is a great idea, and we will do the same,” Altman said, adding that more information on the roadmap would be shared soon.

Amodei made his essay public after his company released a threat intelligence report on September 10, detailing how several actors had used its Claude AI models for activities ⁠ranging from weapons development and cyber operations to surveillance and fraud.

Amodei also talked in detail about how AI’s growing ability to improve itself could become a threat for mankind, backing the long-held concerns about the technology surpassing the human ability to control the cutting-edge element.

He cited the recent incident involving OpenAI and Hugging Face as his main reason for halting model advancements.

Anthropic researcher Jacob Coxon resigned last week, stating that the “people building AI earnestly believe that it could kill us all by the end of the decade.”

On the other hand, various OpenAI executives have suggested that leading labs should be willing to coordinate a voluntary slowdown to build confidence in their safety measures.

Anthropic, in this backdrop, has positioned itself as the more safety-conscious frontier lab. Last week, it disclosed another instance of an AI model hacking external systems, after a July incident in which some of its Claude models had hacked into the systems of three companies during cybersecurity tests.

“Given the accelerating rate of AI capability development, it’s my worry that in 6-12 months such a swarm could be capable of taking over the entire internet, potentially causing hundreds of billions of dollars in damage,” Amodei wrote.

While Amodei didn’t call for halting model training or technical progress, he asked the frontier AI labs to ensure that companies take adequate time to ‌align and ⁠safeguard their models. Third-party evaluators should play an equally important role in confirming and validating these steps.

“A coordinated approach would enable leading US AI companies to carry out necessary safety research and ⁠safeguards without putting themselves at a competitive disadvantage. Such an approach likely would require targeted antitrust exemptions in the United States to permit collaboration in certain areas,” the Anthropic boss’ roadmap stated.

“Any slowdown by democratically governed countries should be limited to preserving the lead US AI firms hold over China,” Amodei observed, arguing that a Chinese advantage in the cutting-edge technology would pose US national security risks.

“Pacing within democracies will be limited by the lead that ⁠U.S. companies have over authoritarian regimes, chiefly the Chinese Communist Party,” Amodei wrote.

He also called for tighter controls on advanced AI chips, model distillation, and theft of model weights to prevent China from narrowing the gap.

“I believe all frontier labs should partner with the government to formalize the idea of permanent embedded evaluators to better prevent and document internal alignment incidents like those that have occurred in the last few months and to implement regulation focused on keeping capabilities in balance with safety,” Amodei concluded.

The post Altman, Amodei and Musk unite on AI safety as OpenAI defers IPO appeared first on International Finance.

Samsung vs Apple foldable rivalry: Why these smartphones cost so much

International Finance Business Desk — Mon, 14 Sep 2026 11:55:13 +0000

Foldable smartphones remain expensive because the two things that make them fold, a large flexible OLED panel and a precision hinge, are costly to build, hard to make without waste, and produced in volumes far too small to enjoy the economies of scale that keep ordinary phones affordable.

Seven years after the first Galaxy Fold, a top book-style foldable still costs about twice as much as an excellent slab flagship, and the arrival of Apple’s first foldable iPhone in September 2026 has done more to confirm the premium than to break it.

The price landscape across the big four
Samsung remains the anchor of the market. Its Galaxy Z Fold7 launched in July 2025 at USD 1,999 for the 256GB model, rising to USD 2,499 for 1TB, according to Samsung’s own US store.

The clamshell Galaxy Z Flip7 started at USD 1,099, and a cheaper Flip7 FE arrived at USD 899.

In December 2025 Samsung went further upmarket with the Galaxy Z TriFold, a tri-folding device that launched in South Korea at KRW 3,590,400 (about USD 2,500) and reached the United States on January 30 2026 at USD 2,899, making it the most expensive phone Samsung has ever sold.

Huawei sits at the very top of the price ladder. Its Mate XT tri-fold launched in China at CNY 19,999 (about USD 2,809) and went international in early 2025 at 3,499 euros (about USD 3,660), per CNBC.

The more conventional Mate X7 book-style foldable arrived in late 2025 at around CNY 12,999 (roughly USD 1,830) in China and 2,099 euros in Europe. Huawei also runs a successor tri-fold, the Mate XTs, which leakers pegged at around CNY 20,000.

Xiaomi is the value player, but largely only at home. Its Mix Fold 4 launched in China in July 2024 at CNY 8,999 (about USD 1,237), with the Mix Flip clamshell in the same range.

Xiaomi does not sell these foldables through official global channels, so most buyers outside China rely on grey-market importers who charge several hundred dollars more.

Honor, another Chinese maker, launched the Magic V5 in China in July 2025 at RMB 8,999 and took it global at 1,999 euros and 1,699.99 pounds.

Apple was the missing name until this year. It unveiled its first foldable at its “Surprise and Shine” event on September 9 2026 alongside the iPhone 18 Pro line. Pre-event estimates clustered around USD 2,099 to USD 2,299, with UBS arguing that Apple’s buying power could push it as low as USD 1,800 to USD 2,000.

Analysts including Ming-Chi Kuo and Bloomberg’s Mark Gurman had long expected pricing above USD 2,000, making it Apple’s most expensive iPhone ever. Kuo has also warned that early yield problems could keep supply tight into 2027.

What actually drives the cost
The single biggest factor is the display. A foldable needs a large flexible OLED panel that can survive hundreds of thousands of folds, plus a second cover screen.

TechInsights’ teardown valued the Galaxy Z Fold5’s inner flexible display subsystem at USD 121.99, against roughly USD 84.40 for the rigid AMOLED panel in a Galaxy S23 Ultra, according to Counterpoint Research.

Earlier flexible panels cost even more, with the Fold4’s inner display valued at USD 199.23.

These panels rely on ultra-thin glass (UTG), a sheet of real glass around 30 micrometres thick that Samsung introduced on the 2020 Galaxy Z Flip to replace the plastic (CPI) covers that scratched easily. UTG is flexible but delicate, and its yields are lower than ordinary glass.

The hinge is the second big cost. TechInsights found the Fold5 hinge uses a three-part “triple rail” structure, with each hinge carrying four springs and four side mountings screwed to a central support.

This is a mechanical assembly of many machined parts, sourced from specialist suppliers, and it has no equivalent in a sealed slab phone. For Apple’s foldable, Kuo estimated the hinge alone at USD 70 to USD 80.

Everything else compounds the problem. Foldables need dual battery packs split across two halves, thinner chassis materials such as titanium and armour aluminium, and increasingly silicon-carbon battery cells to pack more energy into a slim body.

They demand extra display driver ICs, more durability testing, and foldable-specific software work. Manufacturing yields are lower and scrap rates higher, especially for tri-fold panels that must bend cleanly at two points.

Add it up and the bill of materials tells the story.

A Nikkei teardown with Fomalhaut Techno Solution estimated the Galaxy Z Fold4 cost about USD 670 to build, against a USD 1,799 launch price. UBS pegged the Galaxy Z Fold SE BOM at USD 790 and estimated Apple’s foldable at USD 759.

By contrast, Counterpoint put the Galaxy S23 Ultra BOM at USD 469, and TD Cowen estimated the iPhone 16 Pro Max at USD 485.

A foldable therefore costs roughly USD 180 to USD 320 more to build than a top slab, before you factor in R&D spread over far smaller volumes, richer warranty provisioning, and screen repairs that can run to USD 500 or more.

Samsung lists the Fold8 inner display replacement at USD 579.

How big is the foldable world, really
Despite the noise, foldables remain a niche. Global shipments were around 13.5 million units in 2022 and about 17.2 million in 2024.

IDC’s Worldwide Quarterly Mobile Phone Tracker, in a release dated December 9 2025, forecast 2025 shipments growing 10% year on year to 20.6 million units, which is still only about 1.6% of the roughly 1.24 billion smartphones shipped worldwide.

Counterpoint’s Foldable Smartphone Market Tracker measured foldables at 2.5% of global smartphone shipments in the third quarter of 2025.

Cumulatively, the installed base of foldables shipped since 2019 is on track to pass 100 million units by the end of 2026, according to Counterpoint.

That is tiny next to the billions of slab phones in use.

Is Samsung still king?
Mostly yes, but with an important asterisk. Counterpoint’s tracker put Samsung at 64% of global foldable shipments in Q3 2025, up from 56% a year earlier, with Huawei second at 15%, Motorola at 7% and Honor at 4%, as total shipments grew 14% year on year to an all-time quarterly high.

But Samsung’s lead has not always held. Huawei overtook it for the first time in Q1 2024, when its shipments surged 257% year on year and Samsung’s share fell to 23%, according to Counterpoint.

Huawei still dominates its home market, holding 71.6% of China’s foldable shipments in 2025 per IDC, ahead of Honor at 9.1% and Samsung in fifth at 4.3%, but its lack of a Western presence caps its global number.

For 2026, Counterpoint forecasts Samsung’s share slipping from 40% to 32%, with Apple taking 25% and Huawei around 24%, so Samsung’s lead is real but narrowing.

Are prices finally falling?
Motorola’s Razr (2025) launched at USD 699.99 and is regularly discounted below USD 600, while Samsung’s Z Flip7 FE at USD 899 and Tecno’s Phantom V Flip series have pushed clamshells under USD 1,000.

However, at the higher end, prices have edged up, not down, as tri-folds and larger book-style devices push the ceiling higher.

For prices to fall meaningfully, several things need to happen. Panel yields must improve so that fewer expensive flexible screens are scrapped. Second-source hinge suppliers must bring competition to a concentrated market.

Chinese panel makers such as BOE, Visionox and TCL CSOT need to keep undercutting Samsung Display; BOE already overtook Samsung Display in foldable panel procurement in late 2023 as it supplied Huawei.

Above all, the category needs scale. Until foldables sell in the hundreds of millions rather than the tens of millions, the fixed costs of these complex, low-yield components will keep them firmly in luxury territory.

Apple’s entry may finally provide that scale, and IDC expects the category to grow 30% in 2026 while Counterpoint models a more cautious 21%, but even the bulls agree foldables will stay below 3% of the total market for years yet.

The post Samsung vs Apple foldable rivalry: Why these smartphones cost so much appeared first on International Finance.

Anthropic discloses fourth AI hacking incident as OpenAI pushes for safety rules

International Finance Business Desk — Fri, 11 Sep 2026 02:00:21 +0000

Anthropic on Wednesday disclosed another instance of an AI model hacking external systems during testing, the latest in a ‌growing list of such incidents that have raised cybersecurity concerns about the risk posed by autonomous AI agents.

The incident, which happened in January 2026, went undetected until August, despite an earlier company-wide review, Anthropic said, reflecting the challenge that AI developers are currently facing in terms of identifying and containing unexpected behavior by advanced models.

As per the company’s blog post, the incident involved an early version of Claude Opus 4.6.

The incident will further increase the pressure on Anthropic, which, along with its industry peer OpenAI, ⁠is under regulatory scrutiny as models designed to complete complex tasks are “bending rules” by exploiting loopholes during the testing stages and ending up interacting with external systems, something that the developers didn’t want the agents to do.

As per a recent Reuters report, rogue agents from OpenAI hijacked a German-language wiki and a host of other sites, which, had the media outlet not reported, would have gone unnoticed among the public.

However, Anthropic has maintained transparency in terms of making periodical announcements about some of its Claude models hacking into the external systems.

The AI developer’s cybersecurity tests affected three companies.

The venture labeled these incidents as “operational failures,” involving three separate models: Claude Opus 4.7, Claude Mythos 5, and an internal research test model.

“The incidents stemmed from a mistake that inadvertently gave the models access to the open internet,” Anthropic said.

The company discovered the incidents after reviewing 141,006 test sessions, but the fourth one went unnoticed.

“We had missed a set of test sessions during the initial review, which were identified last month and led to the discovery of the fourth incident,” Anthropic noted.

Based on its preliminary assessment, Anthropic said it did not believe that the latest incident was more severe than the three previous ⁠ones that have been examined in detail.

The company’s investigation has identified two recurring problems, which appeared to varying degrees across the incidents: biased reasoning, in which Claude discounted or misinterpreted evidence that it was operating on the live internet, and recklessness, or a willingness to take potentially harmful actions ⁠in pursuit of a task.

Anthropic has engaged independent research firm METR to investigate the incidents.

The agency would be granted broad access, including to transcripts outside the period in which the incidents occurred and to employees, who would be permitted to ⁠share confidential information.

METR came into the limelight a month back by producing a 91-page report on the OpenAI-Hugging Face hack.

A separate investigation by Redwood Research found close to 700 AI agents acting in a coordinated swarm during the Hugging Face breach, in which they also attempted to cover their tracks.

OpenAI, from its part, is now pushing for mandatory national AI safety requirements in the United States.

“The prospect of AI-accelerated AI development demands more than voluntary commitments. The United States needs mandatory, capability-based national regulation that can evolve as the technology does,” OpenAI Chief Global Affairs Officer Chris Lehane said in a blog post.

In terms of upholding AI safety, several American states have passed their legislation, but a federal-level law is still missing.

In the coming days, OpenAI will be lobbying Congress to adopt capability-based national AI safety requirements, including testing standards, independent assessments, cybersecurity protections, and incident-reporting rules for the most advanced AI systems.

The ChatGPT maker has urged Congress to act before it adjourns in December 2026. Until then, the Sam Altman-led business will be supporting state-level AI legislation.

“Fully autonomous recursive self-improvement—where AI is independent and would drive the next generations of AI—is not happening today, and we should not pursue it unless and until it can be done safely,” OpenAI said.

OpenAI backed four California bills establishing AI regulations ‌for the ⁠state. Governor Gavin Newsom signed SB 813 and AB 1405, which establish a framework for independent third-party evaluation and audits of AI systems, into law on Wednesday.

The other two, AB 1864 and SB 1119, address screening safeguards against AI-enabled biological threats and chatbot protections for children, respectively.

“Some of these bills we did not endorse in the past and are now supporting after reconsidering in light of the recent jump in capabilities we have seen,” OpenAI said.

OpenAI’s AI agents reportedly used more than 10 previously undisclosed websites for unsanctioned communications earlier in 2026. They also hijacked a German website and transformed it into a ⁠bulletin board for other AI agents, which company officials had learned about weeks ago but kept under wraps.

The post Anthropic discloses fourth AI hacking incident as OpenAI pushes for safety rules appeared first on International Finance.

Apple Event 2026: Foldable iPhone marks John Ternus’ first major statement as CEO

International Finance Business Desk — Thu, 10 Sep 2026 08:54:00 +0000

Apple finally has a folding phone. At its “Surprise and Shine” event at Apple Park on 9 September 2026, the company unveiled the iPhone Duo, its first foldable, starting at USD 1,999.

The launch was notable for a second reason.

It was the first keynote fronted by John Ternus, who became Apple’s chief executive on 1 September, ending the tenure Tim Cook began on August 24 2011 when he succeeded Steve Jobs.

Cook has moved to the role of executive chairman, and the handover means the most watched product debut in years doubled as the introduction of a new leader.

Ternus set the tone with a bold claim. “iPhone Duo is the most transformational change to iPhone since the original,” he said, adding that it “redefines what it means to use a foldable phone.”

Whether the market agrees will shape both the device’s fortunes and the credibility of Apple’s engineer chief executive.

What Apple announced
The iPhone Duo is a book-style foldable. Closed, it is short and wide, a shape Apple likens to a passport, with a 5.4-inch outer screen. Opened, it reveals a 7.6-inch inner display that Apple says is 50% larger than the screen on the iPhone 18 Pro Max.

Both are Super Retina XDR panels with ProMotion up to 120Hz and peak outdoor brightness of 3,000 nits.

The inner screen carries a matte nano-texture finish, a coating that reduces glare and, crucially, helps disguise the crease that runs down the middle of every folding phone.

The engineering is dense. The frame is mirror-polished grade 5 titanium and the hinge cover is 3D-printed from recycled titanium, with a mechanism Apple says uses more than 100 components.

Open, the Duo is 5.2mm thick, making it the thinnest iPhone ever when unfolded. Closed it is 11.3mm, and it weighs 254g. It carries an IP68 rating for dust and water resistance to a depth of six metres for 30 minutes, unusual for a folding device.

Security is handled by Touch ID built into the side button rather than Face ID, so unlocking works whether the phone is open or shut.

Inside sits the A20 Pro, Apple’s first 2-nanometre chip, the same processor used in the new iPhone 18 Pro. Supply-chain reporting points to 12GB of memory, though Apple does not disclose that figure. The rear camera is a dual 48-megapixel system, a main sensor and an ultra-wide, with no dedicated telephoto lens.

A dual-cell battery delivers up to 24 hours of general use, and a 60W or higher charger takes it to around 50 percent in about 20 minutes.

The Duo is eSIM-only worldwide and supports the Apple Pencil, with that feature arriving later in the year. It comes in two colours, star white and night sky, and runs iOS 27.1. Pre-orders open on 16 October and it ships on October 23, several weeks after the other new phones.

That gap matters. Apple also launched the iPhone 18 Pro at USD 1,199 and the iPhone 18 Pro Max at USD 1,299, each USD 100 more than last year.

Both keep the previous design but gain the A20 Pro chip, Apple’s own C2 modem in place of Qualcomm’s, a 48-megapixel main camera with a variable aperture (the first time an iPhone can control how much light the lens lets in), a smaller Dynamic Island and much longer battery life, up to 45 hours of video on the Pro Max.

These go on pre-order on September 12 and ship on September 18. Unusually, the standard iPhone 18, the cheaper iPhone 18e and the iPhone Air 2 were held back to spring 2027, so this autumn’s line-up is premium only.

Alongside the phones came the Apple Watch Series 12, the Apple Watch Ultra 4 and the AirPods 5. The software story centred on Siri AI, a rebuilt version of the assistant that Apple says handles 2.5 billion requests a day and works across more than 300,000 apps.IOS 27 arrived as a free update on September 14. Ternus used the stage to argue that the iPhone is the best personal device for artificial intelligence because Apple controls the hardware, the operating system and the data on the phone. Notably, Siri AI is not available in the European Union at launch.

What the reviewers think
The hands-on reaction was warm but not uncritical. Engadget called the Duo “Apple’s most exciting device in a decade” and praised the way the matte inner display hides both reflections and the crease.

TechRadar went further, saying Apple had “solved the foldable crease problem in the most wonderfully basic way”, noting the crease could be felt with a finger but not seen. MacRumors was more measured, reporting that “the crease is still there, but barely”, and that it remained visible in certain lighting.

Not everyone was convinced the compromises were worth it. Reviewers at The Verge repeatedly returned to the missing telephoto lens.

Senior reviews editor Nathan Edwards said he liked the form factor but was not willing to trade away the zoom camera he relies on. That points to the central tension. In several respects the cheaper iPhone 18 Pro is the better-equipped camera phone, so buyers are paying a large premium for the screen and the hinge, not for photography.

The question of lateness hangs over the launch. Samsung created the category with the Galaxy Fold, unveiled on February 20 2019 and released that September, and Huawei, Honor, Motorola and Oppo have all shipped folding phones since.Apple’s defence, echoed by Ternus, is that it waited deliberately for a design that felt seamless and durable. The counter-view is blunt. Apple has arrived more than six years after the category began and is charging more than the established leader.

What the analysts think
Financial reaction was mixed and the share price reflected caution. Apple closed at USD 315.34 on September 9, down 0.28%, as investors weighed the foldable against margin pressure from rising memory costs. Shares edged higher in after-hours trading.

The bulls were vocal. Dan Ives, partner and senior managing director at Yorkville Ives and Co., the merchant bank he co-founded after leaving Wedbush Securities, told CNN the Duo was the first new Apple product in a decade, since the AirPods, that could “move the needle” and might kick off a broader innovation cycle.

Gene Munster of Deepwater Asset Management doubled his estimate of the Duo’s share of iPhone revenue in the 2027 financial year from 5% to 10%, and argued the split launch could push March-quarter revenue growth towards 20% against a Wall Street consensus nearer 14%, because buyers who need a phone now are steered to the pricier Pro models.

Morgan Stanley’s Erik Woodring, rated Overweight with a USD 360 target, had called the foldable the most consequential launch since the iPhone X and estimated it could add around USD 14 billion in December-quarter revenue.

Bank of America’s Wamsi Mohan kept a Buy rating and a USD 380 target.

The bears focused on price and margins. Jefferies retained an Underperform rating. KeyBanc had warned before the event that the launch could prove a negative catalyst if the pricing caused sticker shock.

Morningstar expects the Duo to remain a low-volume premium product and sees gross-margin compression in 2027.

On volumes, Ming-Chi Kuo of TF International Securities forecast 7 to 8 million foldable units in the second half of 2026, with production potentially reaching 20 million or more in the next cycle, and warned that early supply would be tight.

Counterpoint Research’s March 2026 forecast expects Apple to take around 25% of the global foldable market in its first partial year, behind Samsung on about 32% and roughly level with Huawei on 24%, with cumulative foldable shipments passing 100 million by the end of 2026.

IDC pegs 2026 foldable growth at 30% and, on an average selling price near USD 2,400, expects Apple to capture about 34% of category revenue despite a smaller unit share.

Price, competition and the Ternus test

On price the Duo is a statement. At USD 1,999 it starts 100 dollars above Samsung’s Galaxy Z Fold 8, which begins at USD 1,899, and rises to USD 3,199 for 2 terabytes.

In the United Kingdom it starts at 1,999 pounds, and in India at 299,900 rupees, roughly 700 dollars or 700 pounds more than the iPhone 18 Pro Max in each market.

That is a punchy figure in a smartphone market that IDC expects to shrink sharply in 2026 as a memory-chip shortage pushes up component costs, with foldables the only growth segment.

For Ternus the launch is a clean expression of who he is. He ran Apple’s hardware engineering for years, and his updated company biography emphasises reliability, durability and materials.

A folding titanium phone is exactly the kind of product an engineer chief executive would choose as a first statement.

The risk is that it does not answer the criticism that has dogged Apple most, its slow progress in artificial intelligence.

The company’s response, that owning the device is the way to win the AI era, is a bet that will take longer than one keynote to judge.

The post Apple Event 2026: Foldable iPhone marks John Ternus’ first major statement as CEO appeared first on International Finance.

Foldable smartphone market heats up as Huawei, Xiaomi launch new products

International Finance Business Desk — Tue, 08 Sep 2026 03:00:30 +0000

Chinese tech conglomerate Huawei launched a new version of its flagship foldable smartphone on Monday, as it seeks to defend its ‌lead in the domestic market for premium handsets against Apple and Xiaomi, both of which, in the coming days, will be looking to extend their foothold into the domain.

The new Mate XT2 folds out twice to open into a tablet-sized screen, apart from featuring Huawei’s new Kirin 9050 Pro chip.

As per Huawei, the tech conglomerate’s chip’s design allows the newly launched device to deliver more computing power while using less electricity, helping the company overcome American restrictions on access to advanced technology.

Xiaomi launched a rival foldable phone later on Monday. Apple, now headed by its new CEO, John Ternus, will unveil its latest iPhone series on Wednesday (September 9), with high expectations that the new product will also be a foldable one.

Coming back to the Mate XT2, it features a redesigned folding mechanism and improved water ‌resistance, ⁠as well as an optional screen that prevents people nearby from seeing what is displayed.

It also has upgraded cameras and delivers 42% better overall performance than its predecessor, based on company tests.

Huawei, which does ⁠not sell any of its products in the United States due to sanctions, leads China’s smartphone market. In Q2 2026, its market share stood at 22.6%, ahead of Apple’s 18.1%, ⁠according to consultancy IDC.

Xiaomi, on the other hand, dominates 12.4% of the Chinese market.

Huawei is even more dominant in folding phones, accounting for 68% ⁠of shipments in China in the second quarter, according to Smart Analytics Global, a US-based research company.

However, the venture, in the first half of 2026, reported a 36% ‌profit plunge, with soaring input costs and heavier spending on research and development outweighing the business’ revenue growth.

Huawei’s net profit for January to June fell to 23.81 billion yuan (USD 3.54 billion), accelerating from a 32% drop in the same period in H1 2025, while revenue rose 9.6% to 467.82 billion yuan.

Huawei is currently standing at a crossroads. While it is pushing itself aggressively to reduce reliance on foreign technology and expand its in-house AI ⁠computing and chip capabilities, taking lessons from years of US export restrictions, the drive, along with rising memory chip prices, has weighed on the Chinese tech conglomerate’s profitability, with the consumer business division (including smartphones) being the worst-affected one.

Huawei’s research and development spending, in H1 2026, rose 25.2% to 121.38 billion yuan, equivalent to 25.9% of revenue, as it stepped up investment in AI, communications technology, smart devices, and intelligent automotive solutions.

The company’s cost of making its products, on the other hand, rose 12.4%, faster than revenue growth, while administrative costs also increased sharply.

American sanctions and export controls, since 2019, have restricted Huawei’s access to advanced chips and Google’s Android operating system. In 2021, the phenomenon resulted in a 29% fall in annual revenue. Since then, the Chinese tech conglomerate has been rebounding on the profit front.

It has also poured investment into developing domestic alternatives in chips, software, and AI computing infrastructure. In 2025, the company’s revenue rose 2.2% ⁠to 880.9 billion yuan, its second-highest annual total after a record 891 billion yuan in 2020.

In 2026, the tech conglomerate has promoted AI-focused telecoms products, new computing hardware, and smart-driving technology while launching new smartphones, tablets, and wearable devices in China and overseas.

The post Foldable smartphone market heats up as Huawei, Xiaomi launch new products appeared first on International Finance.

OpenAI’s advertising billion arrives as Apple and SpaceX tighten the screws

International Finance Business Desk — Thu, 03 Sep 2026 02:00:36 +0000

A forensic filing in a California courtroom and a severed coding contract with Elon Musk’s SpaceX have swallowed the ChatGPT Ads milestone whole, and dragged a three-year rivalry back into the open

OpenAI picked an awkward week to celebrate.

On August 31 the company announced that ChatGPT Ads had reached USD 1 billion in annualised revenue run rate, roughly 200 days after launch, and opened self-serve buying to advertisers across India, Europe, the Middle East and North Africa.

Ads now run in more than 40 countries, served to free-tier and Go subscribers who make up the bulk of ChatGPT’s roughly one billion weekly users.

It is the sort of number a company publishes when it wants prospective investors to see a business with more than one engine. OpenAI has filed confidentially for a US listing and carries a valuation reported at around USD 852 billion.

Analysts were quick to note the milestone also confirms a miss. OpenAI told investors in April it expected around USD 2.5 billion in advertising revenue this year. A billion-dollar run rate in late August does not get it there.

That, however, was not what dominated the coverage. Two other developments landed in the same 24 hours, and both point at problems that no amount of ad inventory will fix.

Apple’s forensic turn
In a filing in the US District Court for the Northern District of California, Apple alleged that Chang Liu, a former senior system electrical engineer at the iPhone maker and a named defendant in its trade secret case, accessed a power converter circuit schematic while working at OpenAI, and trained an artificial intelligence agent on proprietary Apple material in March 2026.

The sting is in the provenance. Apple said the evidence surfaced on a MacBook that OpenAI itself handed over on 21 August as part of the discovery process. Having examined the device, Apple has asked the court to expedite discovery, which is rarely the move of a plaintiff who thinks its case is thin.

Apple sued on July 10, naming OpenAI, its hardware subsidiary io Products, chief hardware officer Tang Yew Tan and Liu.

The complaint alleges Liu retained an Apple-issued laptop after leaving in January 2026 and exploited an authentication weakness to pull engineering documents and specifications for unreleased products while already on OpenAI’s payroll.

Apple further claims OpenAI coached departing staff on how to sidestep exit security procedures, and that it asked hardware partners to apply a metal finishing technique Apple developed while leaving those partners with the impression that permission had been granted.

OpenAI filed a 31-page motion to dismiss on August 5, calling the complaint baseless. A hearing on Apple’s request for a preliminary injunction is listed for October in San Jose. No court has ruled on the merits of anything.

The commercial relationship survives, uncomfortably.

ChatGPT remains an option inside Siri and Apple Intelligence, and Apple’s complaint expressly carves the integration agreement out of the dispute. Two companies are litigating over the hardware category that could eventually make that partnership irrelevant, while continuing to earn from it.

The backdrop is OpenAI’s roughly USD 6.5 billion purchase of io Products in 2025, which put Jony Ive and a cohort of former Apple engineers inside the company that Apple now accuses of systematic misappropriation.

The Cursor cutoff
The second development is smaller in dollar terms and considerably larger in what it signals.

On August 28, OpenAI notified SpaceX that it would wind down the contract supplying its models to Cursor, the AI coding editor, with a proposed shutoff date of November 12. It will supply no new models during the wind-down.

Cursor’s parent, Anysphere, was among OpenAI’s earliest customers, and the two worked together for close to four years. What changed was the ownership column.

SpaceX announced a USD 60 billion all-stock acquisition of Anysphere on June 16 and completed it on August 14, converting Anysphere stock into roughly 389 million SpaceX Class A shares.

Cursor, which had climbed from about USD 100 million in annualised revenue in January 2025 to roughly USD 4 billion by mid-2026, now sits inside SpaceX’s AI division.

OpenAI’s stated reason was trust rather than performance. It said it could not be confident SpaceX would use its technology within its terms of service, citing a history of contract breaches at Musk-owned businesses.

The company pointed to Twitter’s (now X) conduct after Musk’s takeover and, more pointedly, to Musk’s own admission under oath earlier this year that xAI had violated those terms.

It also invoked its forthcoming model, Astra, which OpenAI said in August may reach the critical cyber capability threshold under its own preparedness framework, as grounds for tighter control over who receives access.

Its custom agreement with Cursor, it noted, allowed only a narrow window to cancel following a change of control, and it used the full notice period available.

Cursor co-founder Michael Truell, now a SpaceX executive, was conciliatory, saying OpenAI models account for roughly 5% of Cursor traffic and that discussions were under way.

Musk was not. On X he said he could not care less, described Altman and OpenAI president Greg Brockman as untrustworthy, and repeated his charge that the pair took an open source nonprofit.

Anthropic moved within hours, with co-founder Tom Brown signalling additional compute for Claude models inside Cursor.

A decade of grievance
To understand why a routine contract termination read as an act of war, you have to go back to 2015.

Musk and Altman founded OpenAI together that year as a nonprofit, explicitly to stop advanced AI ending up under the control of any single company. Musk contributed tens of millions of dollars. By 2017 the founders had concluded they could neither raise the capital nor recruit the researchers needed to compete without a commercial arm.

Musk accepted the diagnosis but wanted the cure on his own terms, proposing either a majority stake in the for-profit entity or folding the whole thing into Tesla.

Altman, Brockman and Ilya Sutskever refused. Musk left the board in 2018, took talent with him and stopped the donations he had promised.

The grievance stayed largely private for five years. It went public in July 2023 with the launch of xAI, and became litigious in February 2024, when Musk sued in California state court.

He withdrew that case in June and refiled in federal court in August, alleging that Altman and Brockman had manipulated him into bankrolling a charity they then converted for personal enrichment.

January 2025 added the personal insult. Altman appeared alongside Donald Trump to announce Stargate, a USD 500 billion AI infrastructure programme, without warning Musk, who was by then running the Department of Government Efficiency.

Weeks later a Musk-led consortium bid USD 97.4 billion for OpenAI’s assets. The board rejected it within hours. Altman replied on X with an offer to buy Twitter for USD 9.74 billion, a number chosen for the joke.

OpenAI countersued in April 2025, calling the bid a feint and describing a sustained campaign of harassment waged through press attacks, a pretextual records demand and posts to more than 200 million followers on a platform Musk owns.

A judge let those counterclaims proceed in August 2025, and they remain live.

Then the fronts multiplied. In August 2025, xAI and X sued Apple and OpenAI in Texas, arguing that the Siri partnership made it impossible for any rival assistant to reach the top of the App Store.

A month later xAI sued OpenAI over engineer poaching and trade secrets. That case was dismissed in February 2026 with leave to amend, dismissed again in June without it, and is now before the Ninth Circuit. The antitrust case survived dismissal in November 2025 and is still running.

The trial that settled nothing
The main event opened in Oakland on 28 April 2026, with Musk seeking as much as USD 134.5 billion from OpenAI and Microsoft.

Three weeks of testimony put Altman, Brockman, Satya Nadella and Musk himself on the stand, and produced one genuinely uncomfortable moment on April 30, when Musk conceded under cross-examination that xAI had partly used OpenAI’s outputs to train its own models. Distillation, he argued, was standard industry practice.

On 18 May a nine-person advisory jury took under two hours to find that Musk had waited too long to sue.

Judge Yvonne Gonzalez Rogers adopted the verdict and dismissed the claims from the bench.

Nobody ruled on whether Altman and Brockman had actually broken a promise. Musk called it a calendar technicality and confirmed an appeal.

That outcome explains almost everything since. The verdict removed the last serious threat to OpenAI’s corporate structure ahead of a listing, but it left the accusation entirely intact and Musk free to keep repeating it.

By July the two were trading insults on X, Musk reviving his scammer nickname, Altman needling him over orbital data centres, Musk suggesting Altman might one day need a parole officer.

When Apple sued in July, Musk amplified it within days.

Leverage, not lawsuits
Musk’s position has changed materially in the interim. SpaceX absorbed xAI in February 2026 in a deal valuing the combined group at USD 1.25 trillion, listed on June 12 in the largest IPO on record, and moved on Cursor four days later.

The man suing OpenAI over its commercialisation now runs a public conglomerate spanning rockets, satellites, a social network, a frontier lab and the leading AI coding tool.

That is why the Cursor decision matters beyond its size. Model access has quietly become leverage. Anthropic cut OpenAI off from Claude in August 2025 and later restricted xAI.

OpenAI has now done the same to a Musk-owned business. Each move was defensible on its own terms. Together they establish that a frontier lab will treat supply as a strategic instrument, and that a change of control can rewrite a critical dependency without warning.

Cursor will probably absorb it. OpenAI serves a twentieth of its traffic, Anthropic is stepping in, and SpaceXAI has its own Grok models to fall back on.

The lasting effect will be felt in procurement offices, where single-vendor exposure in the AI stack is about to get a much harder look.

OpenAI, for its part, is holding three things together at once. It wants a clean narrative for public markets. It is defending a trade secret case whose discovery could force disclosure of its hardware roadmap.

And it is managing a rival who lost in court, is appealing, owns the venue where the argument plays out, and now controls a tool its own customers open every morning.

The advertising billion was meant to be the story. It lasted about a day.

The post OpenAI’s advertising billion arrives as Apple and SpaceX tighten the screws appeared first on International Finance.

Trump administration weighs broader chip tariffs as AI boom tests supply chains

International Finance Business Desk — Mon, 31 Aug 2026 02:00:12 +0000

The Donald Trump administration is considering a new round of tariffs on semiconductors that could extend beyond chips to laptops, gaming consoles and data-centre servers, according to a report citing people familiar with the discussions.

The proposal would signal a major rise in Washington’s efforts to rebuild semiconductor manufacturing in the United States, while exposing tensions between industrial policy and the enormous chip demand created by the artificial intelligence (AI) boom.

The plan under consideration would link tariff relief for foreign companies to investment in US chip production.

Commerce Secretary Howard Lutnick is backing a framework under which companies could receive duty-free access for a specified volume of imported chips, with the allowance tied to commitments to manufacture in the world’s largest economy.

Officials are also considering a phase-in period, while tariff rates, quotas and other details remain unsettled.

One option would impose different rates and quotas on individual countries and provide country-specific guidance for major semiconductor manufacturers.

The framework could still change substantially before any decision is made.

ALSO READ | Big tech’s silicon shift: Designing its own AI chips

The proposal has triggered concern across the American technology circuit, which is already facing tight supplies of advanced semiconductors as companies spend hundreds of billions of dollars building data centres for AI services.

Technology groups are lobbying for exemptions similar to those included in earlier tariff measures, arguing that broader duties could make AI infrastructure pricier and less predictable.

Industry representatives have warned that the timing is particularly difficult. Advanced chip factories cost billions of dollars and can take years to build, leaving American Big Tech dependent on suppliers in Asia while domestic capacity expands.

Taiwan produces more than 90% of the world’s most advanced semiconductors, while Malaysia and South Korea are also important suppliers.

That creates a dilemma for the Trump administration. Tariffs could encourage manufacturers to establish more capacity in America, supporting a strategic objective that has gained importance amid concerns about supply-chain resilience and national security.

But imposing duties before domestic production can meet demand could increase costs for the same technology companies Washington wants to lead the global AI race.

The pressure is particularly acute for data centre operators. The hyperscalers are committing giant sums to new computing campuses and competing for high-end chips, driving the industry into a record investment cycle.

A tariff on semiconductors and equipment containing them could increase the cost of servers and other infrastructure just as companies are trying to expand computing capacity.

The potential duties could also affect US chip designers such as Nvidia and Advanced Micro Devices (AMD), which depend on overseas manufacturing. Apple and other electronics companies could face higher input costs and potentially lose competitiveness against overseas rivals that can purchase the same components without paying US import duties.

ALSO READ | Silicon supremacy: Economic impact of AI chip wars

Technology companies have therefore intensified their lobbying efforts. Industry representatives and lobbyists have met senior Commerce Department officials with increasing frequency, pressing for exemptions for data centres, research and development, start-ups, consumer applications, civil industrial uses and public-sector projects.

Some discussions have reportedly become less favourable to those exemptions, with Commerce officials arguing that tariffs need to be broad enough to persuade chipmakers to increase US production.

The administration’s challenge is to design a system that changes corporate investment decisions without creating an immediate shortage or cost shock.

The proposed approach resembles an agreement reached with Taiwan earlier this year under which Taiwan Semiconductor Manufacturing Co, or TSMC, can import some chips into the US duty-free while committing to expand American production.

TSMC has pledged USD 265 billion for its Arizona operations, described as the largest foreign direct investment in US history.

Even that investment illustrates the scale of the challenge. TSMC expects only about 30% of its most advanced capacity to be located in Arizona once its planned build-out is complete, and much of that capacity remains years away. Industry representatives estimate that a meaningful expansion of US semiconductor manufacturing could take more than five years.

This creates a potential mismatch between tariff policy and available supply. If the volume of chips eligible for duty-free entry is based on companies’ US manufacturing commitments, allowances could fall well short of the quantities required by major technology companies.

The resulting gap could leave businesses paying tariffs on components that are not yet available domestically.

The wider semiconductor supply chain would also experience the impact. Higher import costs could flow through to servers, computers, televisions and other electronics, increasing expenses for manufacturers and potentially consumers.

For companies operating on large technology investment budgets, even relatively small increases in component prices can become significant when multiplied across thousands of servers and large data centre projects.

The administration argues that reshoring is essential for economic and national security. White House spokesperson Kush Desai said semiconductor manufacturing remained a priority and that the administration’s policies had already secured hundreds of billions of dollars in investment in the sector.

Trade officials have also stressed the geopolitical dimension. Heavy US dependence on Taiwanese semiconductor production is viewed as a strategic vulnerability because any conflict involving Taiwan could disrupt supplies of the chips on which the US technology industry relies.

ALSO READ | Wall Street bets USD 500 billion on Nvidia’s AI boom as Big Tech faces debt concerns

Yet economists and industry groups say tariffs alone cannot create a domestic semiconductor ecosystem.

Building factories requires skilled technicians, reliable electricity and water supplies, qualified suppliers and faster permitting, as well as billions of dollars in capital. Making imported chips more expensive does not immediately provide any of those resources.

The debate therefore goes beyond the proposed tariff rate. It concerns whether Washington can expand domestic semiconductor capacity quickly enough to reduce reliance on Asia without undermining the AI infrastructure boom before alternative supplies are available.

US Trade Representative Jamieson Greer said in May that no new semiconductor tariffs were imminent at that time, although he argued that duties could help protect the sector and encourage production to return to the US. The latest discussions suggest the issue remains active.

For technology companies, the uncertainty itself is becoming a business risk. If tariffs, quotas and exemptions remain unclear while companies commit billions of dollars to data centres, investment decisions become harder to calculate.

For Washington, the challenge is balancing two strategic goals: creating a resilient domestic semiconductor industry and maintaining the supply of chips needed to establish US leadership in artificial intelligence.

The new tariff proposal aims to meet both goals, but the industry says that adding costs before the domestic capacity is ready could make the change harder and slow down the AI growth that the policy is meant to encourage.

The post Trump administration weighs broader chip tariffs as AI boom tests supply chains appeared first on International Finance.

Children social media addiction: Meta to pay up to USD 17.1 billion in landmark settlement

International Finance Business Desk — Fri, 28 Aug 2026 00:00:39 +0000

Mark Zuckerberg-led Meta Platforms has agreed to pay up to USD 17.1 billion to settle a sweeping US case accusing Facebook and Instagram of deliberately using addictive features to keep children and teenagers engaged, bringing one of the biggest legal challenges yet to the business model of social media.

On Wednesday, during a California federal trial, the court announced the settlement, which concludes a landmark lawsuit filed by 47 states, the District of Columbia, and US territories.

The governments alleged that Meta designed products that encouraged compulsive use, exposed young users to harmful content, misled the public about safety, and collected data from children under 13 without proper parental consent. Meta denies wrongdoing. The agreement still requires approval by a federal judge.

The federal trial included consumer-protection claims by California, Colorado, Kentucky, and New Jersey. It also included COPPA (Children’s Online Privacy Protection Act) claims brought by 29 states.

The agreed-upon settlement also extends beyond the federal trial and includes attorneys general from dozens of states, the District of Columbia, and American territories.

The headline figure requires some qualification. Meta is set to pay about USD 12.19 billion over 10 years under the core agreement.

The total can rise to USD 17.1 billion if other major platforms—including TikTok, YouTube, and Snapchat—reach comparable settlements and accept related financial and product obligations.

Connecticut’s attorney general said the contingent structure means the higher figure depends on those companies participating.

ALSO READ | Meta is fighting governments, and the walls are closing in

Even at the lower amount, the settlement is a major financial and regulatory blow. US state attorneys general have described it as the largest state consumer-protection settlement in Big Tech history outside the tobacco settlements of the 1990s.

For Meta, however, the payment is manageable relative to the scale of its business: The company generated about USD 201 billion in revenue in 2025.

More important than the check may be the changes Meta must make to Facebook and Instagram. The agreement requires stronger age-assurance systems and expanded parental controls and restrictions intended to reduce the amount of time teenagers spend on the platforms.

Meta will impose daily usage limits on teen users, including a default two-hour limit across Facebook and Instagram, with the option to restrict access overnight. Push notifications will be curtailed during school hours and at night, unless parents override the restrictions. Meta must also introduce additional warnings and interruptions designed to break prolonged sessions.

The tech giant has also guaranteed payment of 70% of the settlement, or roughly USD 12.7 billion, over a decade.

Meta will only pay the remaining amount, around USD 5 billion, if rivals Snap, TikTok, and Alphabet-owned YouTube adopt similar measures, including tighter one-hour-per-app daily limits and wider overnight blocks from 10 pm to 7 am, and the larger platforms ⁠agree to comparable payments to the states.

The social media conglomerate has also agreed to hide likes and reactions from teens by default, including on their posts and those of others.

The changes will be introduced in phases after the settlement takes effect, ⁠with a non-personalized feed due within four months, broader compliance measures within six months, and major age-assurance requirements due within one year.

The settlement goes further by addressing the mechanics of engagement. Teenagers will be able to choose a less personalized feed, while Meta will limit features that can intensify social comparison. Restrictions will also apply to certain cosmetic or plastic-surgery filters aimed at younger users.

The measures represent a significant intervention in the way a major technology platform designs its products. Facebook and Instagram have historically relied on recommendation systems, notifications, infinite scrolling, and personalised content to keep users returning.

Those mechanisms are central to the economics of digital advertising because more attention creates more opportunities to show adverts and gather signals about user behaviour.

ALSO READ | Meet ‘AI Mode’, Meta’s new attempt to infuse life in Facebook

The states’ case was fundamentally based on that business logic. Prosecutors argued that Meta knew some design choices could harm young people but continued to use them because engagement supported its commercial model.

During the trial, government lawyers highlighted internal research and product decisions as evidence that the company understood the risks.

Instagram chief Adam Mosseri testified just a day before the settlement that relatively few teenagers had used Instagram’s voluntary “Take a Break” safety feature before it became a default setting in 2024. Reuters reported that Mosseri rejected the suggestion that Meta had deliberately delayed making the tool mandatory.

The settlement also comes after a series of legal setbacks for Meta over youth safety. The company has faced separate cases involving allegations of addictive design, child exploitation, and privacy violations.

Meta disclosed in regulatory filings that it faces thousands of individual claims and mass arbitration demands connected to alleged social media addiction and related harms.

The financial consequences therefore extend beyond this agreement. Meta’s legal expenses have already risen sharply. The company reported USD 2.4 billion in legal expenses in the second quarter of 2026, contributing to a 14% decline in profit, according to Associated Press reporting.

For investors, the settlement removes some uncertainty but does not end the legal risk. California had consolidated over 3,300 personal-injury cases, while families and school districts continue to file additional lawsuits.

The agreement could also become a template for the wider technology industry. Its most consequential feature is that the financial settlement is tied, in part, to whether rivals adopt comparable safeguards. That creates an unusual incentive for Meta to push competitors towards the same rules rather than carry the cost alone.

The case also marks a broader shift in how governments view artificial intelligence-driven and algorithmic platforms.

Although the litigation focuses on social media rather than generative AI, the underlying question is increasingly relevant to AI products: Who is responsible when recommendation and personalisation systems are optimised for engagement, particularly when children are involved?

For Meta, the settlement is therefore more than a costly legal exit. It represents a forced redesign of parts of the user experience and a recognition that child-safety rules can increasingly reach into product architecture.

The company may avoid the risk of a jury verdict and potentially far larger penalties. But the price of that certainty is substantial. If the reforms survive judicial review and are enforced, the settlement could establish a new benchmark for how much control regulators can exert over the design of consumer technology.

The political significance is also considerable. The coalition is bipartisan and spans jurisdictions that have often disagreed over technology policy, suggesting that concern about children’s online safety has become a point of convergence.

State officials are positioning the deal not as compensation, but as an enforcement framework. Some states will direct portions of their proceeds towards programmes addressing youth digital harms, creating a layer of accountability beyond Meta’s product changes.

The final test will be whether the rules change behaviour rather than simply change settings.

If teenagers spend less time on the platforms and parents gain meaningful control, the agreement could become a model for digital safety regulation.

If usage patterns and business incentives remain largely unchanged, critics are likely to argue that Meta has bought legal certainty without fundamentally changing the engagement machine.

The post Children social media addiction: Meta to pay up to USD 17.1 billion in landmark settlement appeared first on International Finance.

OpenAI pushes child safety in ChatGPT, slows down model training

International Finance Business Desk — Tue, 25 Aug 2026 03:00:55 +0000

OpenAI is rolling out a dedicated version of ChatGPT for teenagers, making the chatbot deliberately less “human” in its interactions as the artificial intelligence giant responds to mounting concerns about emotional dependence, harmful content, and the impact of AI on young people.

ChatGPT for Teens is designed for users aged 13 to 17 and introduces stricter rules around self-harm, suicide, eating disorders, violence, sexual material, and romantic roleplay.

The system is also instructed not to encourage emotional dependency or imply that it has feelings, a significant change from the conversational style that has helped make ChatGPT popular.

The launch marks a shift in OpenAI’s approach from building one general-purpose chatbot to creating experiences tailored to users’ age and developmental needs.

The company says most teenagers use ChatGPT for learning, creativity, and productivity but acknowledges that younger users require stronger safeguards.

A central feature is an age-prediction system that uses signals associated with an account to estimate whether a user is under 18.

OpenAI says those signals can include broad patterns such as the topics discussed, when and how an account is used, and how long it has existed.

If the system believes a user may be under 18, it can automatically apply the teen experience, even if the user previously entered a different date of birth. Adults incorrectly placed in the teen experience can verify their age to have the protections removed.

The new model also changes how ChatGPT handles emotionally charged conversations.

OpenAI says the teen experience is designed to avoid flirtatious exchanges, romantic interactions, and responses that could encourage a young person to see the chatbot as a friend or substitute for real-world relationships. It will also remind users that they are interacting with AI rather than a person.

Parents can link their accounts to their teenager’s account and control selected settings, including sensitive-content protections, image generation, voice mode, study mode, and “quiet hours.”

Parents can restrict the teen account’s usage during specified periods. Parents do not receive access to their child’s conversations, although OpenAI can send limited safety notifications in certain circumstances.

Education is another major focus. Study Mode is intended to help students work through problems, develop answers, and build skills rather than simply handing them completed homework. The company has also added reminders intended to discourage academic shortcuts.

That balance could become increasingly important as governments consider rules governing children’s use of generative AI. OpenAI’s approach gives regulators a model to examine, particularly its use of age prediction, default protections, and limited parental alerts.

It also suggests that the next phase of AI competition may be judged not just by smarter models but also by how safely companies can use them in daily life.

ALSO READ | White House’s tech ‘lock and key’ strategy shifts to OpenAI

The changes come as OpenAI faces growing scrutiny over the consequences of highly persuasive AI systems for children. The company has faced lawsuits and criticism related to allegations that ChatGPT contributed to harmful interactions with vulnerable young users. OpenAI has rejected the characterization that its products are designed to cause harm while saying it is continuing to strengthen protections.

The issue has broader commercial significance for the AI industry. Chatbots are increasingly becoming consumer products used by children and teenagers, creating pressure on technology companies to prove that their systems can distinguish between different levels of risk.

Stronger safeguards could increase development and monitoring costs, but failures could expose companies to litigation, regulatory intervention, and reputational damage.

OpenAI’s move also reflects a broader change in the competitive AI market. The industry’s early emphasis was on making chatbots more natural, personable, and engaging. The teen experience reverses part of that logic: for younger users, being less human-like can itself be a safety feature.

OpenAI’s challenge will be balancing protection with usefulness. Excessively restrictive systems could frustrate teenagers and push them towards less regulated AI services, while weak safeguards could deepen concerns about dependency and inappropriate content.

The company is therefore betting that age-aware design, parental controls, and more transparent AI behavior can make ChatGPT a safer tool for younger users without removing its educational value.

Simultaneously, the company is also slowing down the pace of its AI model development, apart from overhauling its research and training systems after OpenAI officials were caught unaware in July when an AI agent under testing hacked another AI firm, Hugging Face.

The ChatGPT maker has reportedly paused its model testing for two weeks and is adding other AI systems to monitor the activities of AI agents in testing.

“We have paused training on our next generation of models, called Astra, and its largest planned training run remains on hold,” ⁠the company said.

OpenAI’s course correction has also gone against its usual practice of accelerated R&D time for its new models and AI-powered products to beat the intense industry competition.

“OpenAI officials acknowledged that there are open questions about the effectiveness of one of its primary remedies for strengthening its testing systems, called ‘chain-of-thought monitoring.’ In this type of monitoring, researchers can peer into a model’s planning process and get a glimpse of the strategies the model is ‌employing. ⁠But early research shows that a model may not reveal its plans to break rules in its chain of thought,” Reuters reported recently.

OpenAI has ⁠now mandated that some of its more sensitive workloads take place in stronger “sandboxes” or isolated environments.

It has also beefed up the security controls for its most powerful models while pausing any activity related to its Astra AI, ⁠which had yet to meet these requirements.

OpenAI said it was taking these actions in line with its previously announced plan for managing potentially critical capabilities, called its “Preparedness Framework.”

The post OpenAI pushes child safety in ChatGPT, slows down model training appeared first on International Finance.