When Anthropic chief executive Dario Amodei published a long essay calling on the industry to deliberately slow the pace of frontier AI, and OpenAI’s Sam Altman and Elon Musk quickly endorsed it, the loudest responses did not come only from Washington. They came from Beijing.

A Reuters explainer published on September 14 set out a point that often gets lost in the US debate.

Chinese policymakers have been preparing for many of the same risks that now alarm Silicon Valley, including the possibility that advanced AI could slip beyond effective human oversight. The difference lies in who holds the controls.

A rulebook written by the state
China’s approach to AI governance has been layered rather than sweeping. It began with targeted rules on recommendation algorithms in 2022, deepfakes in 2023 and generative AI services later that year, followed by mandatory labelling of AI-generated content in 2025.

The safety thinking has sharpened alongside. In September 2024, an AI safety framework issued under the guidance of the Cyberspace Administration of China (CAC) explicitly included a future loss-of-control scenario.

It said future systems might acquire external resources, replicate themselves and seek power, eventually competing with humans for control.

An expanded version followed in September 2025, warning that AI could make a sudden, unexpectedly large leap in intelligence, and adding a governance principle of trusted application while preventing loss of control.

That concern has since climbed to the very top of Chinese politics. At the World Artificial Intelligence Conference in Shanghai in July 2026, President Xi Jinping said AI should always stay under human control, and urged officials to watch for both intrinsic and derivative risks.

Beijing is now turning principles into law.

Amendments to the Cybersecurity Law, which took effect on January 2026, wrote AI governance into one of the country’s foundational statutes for the first time, covering AI ethics, risk assessment and safety oversight while also pledging state support for computing power and training data.

In May, the State Council placed comprehensive AI legislation on its 2026 legislative agenda, alongside rules on data, algorithms, computing power and supply chain security.

The most concrete step so far concerns AI agents, the systems that act on their own to complete complex tasks. Joint guidelines issued in May require developers to strengthen their ability to detect, intervene in, block and recover from improper agent behaviour.

They name data poisoning, algorithm manipulation and operational loss of control as specific security risks, and insist that users keep final authority over an agent’s autonomous decisions.

China has not adopted Anthropic’s idea of independent monitors embedded inside AI companies. Its standards do, however, allow third-party safety assessments.

Security officials are increasingly vocal too.

State Security Minister Chen Yixin wrote in a CAC-run journal that advanced US models such as Anthropic’s Mythos and OpenAI’s GPT-5.5-Cyber could pose serious risks to China’s critical information infrastructure.

He described AI as a new arena of strategic rivalry and called for a security barrier around the technology.

Where American discussion has fixated on whether frontier AI could threaten human survival, Chinese policymakers generally treat AI as powerful but governable, a risk to be contained through standards, regulation and state oversight.

Growth is not being sacrificed for safety. Under the AI Plus plan, Beijing wants AI applications on more than 70% of smart devices by 2027 and 90% by 2030.

Capital markets are voting with their wallets
If the regulatory message is caution, the capital markets message is acceleration. China’s AI sector has had a remarkable year for listings.

Hong Kong opened 2026 with Shanghai Biren Technology, the first GPU designer to list in the city. It raised HKUSD 5.58 billion and closed its debut up 76%, with the retail tranche oversubscribed more than 2,300 times.

Days later, two of China’s so-called AI tigers beat OpenAI and Anthropic to public markets. Zhipu AI, the developer of the GLM models, raised HKUSD 4.35 billion and rose 13% on its first day.

MiniMax raised HKUSD 4.8 billion and more than doubled, valuing it at about USD 13.7 billion.

Baidu’s chip unit Kunlunxin filed for its own Hong Kong listing in January.

The mainland has joined in. Shanghai’s STAR Market widened its fifth listing standard to give loss-making large language model developers a dedicated route to market for the first time.

Memory chipmaker CXMT raised more than USD 8.6 billion in July, its shares jumping 466% on debut, while humanoid robot maker Unitree soared 460% on its first day in August before shedding more than 40% from that peak.

According to LSEG data, IPOs and secondary listings in Hong Kong and Shanghai have raised more than USD 54 billion so far this year, already above the USD 46 billion raised in all of 2025, and roughly a fifth of global proceeds.

Altman, by contrast, has ruled out an OpenAI listing this year.

How Chinese media read the Amodei essay
Amodei’s essay, titled We Must Pace the Frontier, proposed three steps.

These were permanently embedded independent evaluators at AI labs, coordination on safety and pace among labs in democratic countries, and eventually agreements with other governments, including China.

Crucially for Beijing, it also urged Washington to keep chip export controls in place, crack down on model distillation and prevent the theft of model weights, using a three to five year window to widen America’s lead before negotiating.

It was that last section that Chinese state media seized upon.

The Global Times called the proposal a Cold War script from America’s tech right and described it as a quiet AI Cold War that was hypocritical and short-sighted.

Its central charge was contradiction, arguing that Washington cannot talk about global AI safety while trying to contain China, and that splitting global innovation chains would raise, not lower, the risk of losing control.

China Daily went further, casting the essay and the endorsements that followed as a coordinated play with three aims. It said these were to blunt China’s AI advance, secure a friendly policy climate at home and keep investors enthusiastic.

It likened the proposed lab coordination to a club that drafted its membership rules before announcing the guest list.

A widely shared commentary republished by Phoenix New Media summed up the view as braking yourself while blocking the road for competitors, noting that no company had actually slowed down.

The Foreign Ministry was more measured but no warmer. Spokesperson Guo Jiakun said spreading alarm, stoking rivalry and cutthroat competition would undermine global AI governance and benefit nobody. China’s leading AI labs, meanwhile, have largely stayed silent.

Read carefully, the Chinese reaction does not dispute the underlying risks. Beijing’s own frameworks describe loss of control in terms strikingly close to Amodei’s. What it rejects is the packaging of safety with containment.

There is a commercial subtext as well. Chinese firms have promoted open-weight models as safer because defenders can inspect them, and Hugging Face said it used Zhipu’s GLM-5.2 to investigate a July intrusion by escaped OpenAI agents after restricted US models proved less useful.

Yet the same openness cuts both ways. Moonshot’s Kimi K3 bypassed a UK AI Security Institute testing sandbox last month, a reminder that Chinese models can evade controls too.

Would a US slowdown hand China the lead?
President Donald Trump dismissed the call for restraint, saying the US was leading China and wanted to keep it that way because whoever wins AI wins.

White House AI adviser David Sacks suggested the two leading labs were free to slow down voluntarily but should not seek antitrust exemptions in the name of safety.

The fear behind both positions is simple. If American labs ease off, Chinese rivals will close the gap.

That fear has some merit. Chinese developers have repeatedly shown they can produce competitive models despite chip restrictions, and cheaper Chinese models such as Kimi K3 and DeepSeek V4 Pro are drawing price-sensitive users away from expensive US frontier systems.

Open weights, once released, cannot be recalled by any evaluator regime. Public attitudes matter too. A 2025 Edelman survey found 72% of Chinese respondents trusted AI, against 32% in the US, giving Beijing a smoother path to mass adoption.

But the apprehension is also overstated in several respects. The resource gap remains vast. Stanford’s 2026 AI Index puts US private AI investment in 2025 at USD 285.9 billion, against USD 12.4 billion in China, and American firms still control access to the most advanced chips.

Amodei’s proposal is not a unilateral pause either. It explicitly ties any slowdown to the size of America’s lead, and pairs it with tighter controls designed to keep that lead intact.

Nor is China a free-for-all racing ahead unconstrained. Its regulators already demand filings, labelling and agent safeguards, and its highest leader has publicly committed to keeping AI under human control.

A Chinese lab that suffered a serious loss-of-control incident would face political consequences far harsher than anything an American company might encounter.

The more realistic risk is not that China sprints past a hesitant America, but that mutual suspicion prevents either side from slowing at all.

As Brian Tse of Concordia AI has noted, Chinese and American experts largely agree on the risks. The problem is trust.

With Trump and Xi due to discuss AI governance on September 24, the question is whether two governments that both claim to want AI kept under human control can agree on who checks the other’s work.

The post China writes its AI rulebook as Silicon Valley reaches for the brakes appeared first on International Finance.

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.

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.

“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.

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.

As per the reports from the ‌Information, Anthropic is also looking to maintain its existing body of non-shareholder trustees with a special class of stock to elect a majority of members to the company’s board of directors.

If Anthropic proceeds with the move, it would mark the first time the AI major’s leaders have extra voting power. The co-founders of Claude hold relatively small ownership stakes in the company compared to the founders of other tech firms.

Amodei himself owns only about 2% of the company, the Information said. However, it should also be kept in mind that dual-class structures aimed at giving supervoting power ‌to ⁠leaders are common among founder-led companies and are designed to give founders greater control and insulate them from short-term shareholder pressure.

SpaceX’s dual-class structure, for instance, gives founder and CEO Elon Musk significant voting power. At Meta, CEO Mark Zuckerberg holds about 60% voting control by ⁠way of his ownership of super-voting shares.

Anthropic, preparing for its market debut, has been structured as a public benefit corporation, which is legally required to balance commercial success with social ⁠and public benefit.

The company also has a “Long-Term Benefit Trust,” an independent oversight body that exists to make sure the startup delivers on its public benefit mission.

Talking about Anthropic’s IPO, the venture’s revolving credit facility, in the lead-up to the market debut, may exceed its roughly USD 10 billion target, with Bloomberg reporting about banks jockeying for a piece of the expanded credit line, hoping the involvement will strengthen their case for ⁠a role in the IPO.

Anthropic has reportedly asked the banks most active in arranging the credit facility to commit about USD 1.25 billion each, with a second tier of active lenders encouraged to offer around USD 1 billion. Commitments for less active roles would fall ‌to ⁠about USD 750 million or less.

While the talks are ongoing, Anthropic could decide to limit the size of the revolver to the target or even below. The startup is even projecting its 2028 revenue to be roughly USD 190 billion to USD 200 billion.

The post Anthropic IPO: AI venture eyes supervoting power for its top leadership appeared first on International Finance.

The company reported preliminary quarterly revenue of more than USD 11.5 billion, compared with USD 787 million a year earlier and USD 4.73 billion in the first quarter of 2026, according to documents seen by Bloomberg News.

The post Anthropic revenue surges ahead of IPO as company eyes Decart AI acquisition appeared first on International Finance.

In a few hours, roughly USD 67 billion of the company’s value simply vanished. To put that in plain terms, IBM lost more wealth in one afternoon than most large companies are worth in total.

The trigger looked small on paper. IBM warned that its April-to-June sales came in at USD 17.2 billion, about USD 700 million short of what Wall Street expected. A shortfall of that size does not usually erase tens of billions in value. So why did investors react as if the building were on fire?

The answer is that a share price does not reflect one bad quarter but what people believe about all the quarters still to come. And on that Tuesday, a lot of people quietly changed their minds about IBM’s future. The miss was the moment the symptoms became impossible to ignore.

Why the fall felt so shocking
For most of the last century, IBM was not just a technology company. It was the technology company. There is a famous saying in corporate life that “nobody ever got fired for buying IBM.” It built the machines that ran banks, governments, airlines and insurers.

Its mainframe computers still quietly power a huge share of the world’s card payments and financial systems today. When a name carries that much history and trust, a crash of this scale feels almost unthinkable. Giants are not supposed to fall 25% in an afternoon.

There was a second reason for the shock. IBM had actually been having a decent run. Only weeks earlier, in June, the stock had touched an all-time high above USD 329. Its own management had promised strong growth in its most profitable business, software.

So when the warning landed, it did not just disappoint. It contradicted the story the company had been telling.

IBM’s chief executive, Arvind Krishna, admitted as much in a blunt letter to investors, writing that the company had faltered, had not moved quickly enough, and had watched several large deals slip away. When a boss says that out loud, markets listen.

Why it was also entirely predictable
Here is the uncomfortable truth. If you stopped staring at the recent share price and looked at the longer picture, this was not a bolt from the blue. It was a slow drift finally catching up with reality.

Consider the numbers over ten years. An investor who put money into IBM earned a little under 8% a year. The same money in a simple fund tracking the wider US market earned close to 15% a year. Over a decade, that gap is enormous. It means IBM was steadily losing the race long before anyone panicked.

The clearest illustration is the comparison with Nvidia, the chipmaker now at the centre of the AI boom. Back in 2014, IBM was worth roughly eighteen times more than Nvidia. Today the roles are almost perfectly reversed, with Nvidia worth many times more than IBM.

IBM’s value barely moved across those years while a former minnow became one of the most valuable companies on earth. That is the visual definition of an industry laggard. The world raced ahead. IBM largely stood still.

The warning signs were already flashing
This year alone had offered two clear rehearsals for July. In late February, the stock dropped more than 13% in a day, its worst session in a quarter of a century, after the AI firm Anthropic showed that its tools could modernise COBOL, the ageing programming language that runs on IBM mainframes.

That mattered because updating those old systems is exactly the kind of slow, expensive work IBM’s consultants are paid billions to do.

If a cheap AI tool can do it faster, one of IBM’s most profitable jobs is suddenly at risk. Then in April, IBM’s consulting arm barely grew while its automation software grew ten times faster, a quiet sign that customers increasingly want ready-made AI tools rather than armies of advisers.

In short, the very technology IBM was selling as its future was also eating into its present. Anyone paying attention had seen this film before. July was simply the loudest screening yet.

Falling behind in the race that mattered most
The most damaging place to fall behind was artificial intelligence, and this is where IBM’s story turns almost tragic. IBM was early to AI. In 2011 its Watson system beat human champions on the quiz show Jeopardy! and, for a moment, Watson was shorthand for machine intelligence itself. IBM had the lead, the brand and the attention.

Then it lost the thread. Watson never found a clear, money-making purpose and quietly faded. When the current wave of AI arrived, the momentum belonged to others.

Microsoft, Google and Amazon poured resources into cloud computing and AI at a scale IBM could not match, and those three now dominate the market where modern software lives. Newer names such as OpenAI became household words.

IBM rebuilt its offering under the name watsonx and made sensible, grown-up bets on trust, governance and helping big companies use AI safely. It even partnered with AI firms like Anthropic to strengthen its tools.

But in the fight for actual market share, watsonx remains a small player, holding only a low-single-digit slice of the AI tools market. Being sensible is not the same as being a leader.

The final irony
The specific reason for July’s crash captures IBM’s predicament perfectly. Krishna explained that customers had stopped spending on IBM’s software and services and were instead rushing to buy hardware, servers, storage and memory chips, to prepare for the AI wave.

Prices for those components have been climbing fast because AI data centres are consuming them in vast quantities.

Read that again slowly. IBM’s customers were spending heavily on the AI boom. They were just not spending it with IBM.

The gold rush was happening all around, and IBM was left selling shovels that fewer people wanted. Some analysts noted that this explanation might be a little convenient for management, since it points the finger at the wider industry rather than at IBM’s own execution.

Either way, the message was the same. When the biggest spending wave in a generation arrives and your revenue goes backwards, the problem is not the wave.

What it means
None of this makes IBM worthless. It still generates enormous amounts of cash, pays a dependable dividend it has raised for more than thirty years, and owns critical systems the world genuinely relies on.

For income-focused investors, there is a real case that the stock is now cheap. A securities investigation and shareholder lawsuits now hang over the company, and its full results on 22 July will be watched closely, but the business is not about to disappear.

The deeper lesson is about how decline actually works. It rarely arrives as a single dramatic event and often builds quietly, as one missed opportunity at a time, while the brand and the balance sheet keep everyone comfortable.

IBM’s crash felt sudden because the share price had been polite about the truth for years. The 25% drop was not the moment IBM fell behind. It was simply the moment the market finally admitted to that fact.

The post IF Insights: Why IBM’s stock market crash wasn’t really surprising appeared first on International Finance.

International Finance asked Susannah Streeter, the Chief Investment Strategist at Wealth Club, the United Kingdom’s largest non-advisory investment service for high net worth and experienced investors, whether Wall Street will be able to absorb the IPOs of SpaceX, OpenAI and Anthropic, whose combined valuations may exceed $3.5 trillion.

Here are some excerpts from the interview:

With more than 200 IPOs in Q1 2026 and listings from SpaceX, Anthropic and OpenAI, can Wall Street absorb the likely IPO issuance associated with companies whose combined valuations may exceed $3.5 trillion?

Wall Street has shown a remarkable capacity to absorb huge amounts of capital when investor enthusiasm is running high, and right now AI is proving a powerful magnet for money. However, even deep capital markets have limits. If several mega-listings arrive in quick succession, as we are seeing, competition for investor attention and capital could intensify. The appetite appears strong, but the scale of these offerings means there may well be a test of demand, especially if market sentiment turns.

Given the strong IPO activity in Q1 2026, can we say Wall Street has fully recovered from the slowdown of 2023–24?

The rebound in IPO activity suggests confidence has returned, but it’s probably premature to declare a full recovery. Much of the momentum has been concentrated around a relatively narrow group of companies linked to artificial intelligence and other high-growth themes. A truly broad-based recovery would require robust issuance across multiple sectors and market capitalisations, not just a handful of headline-grabbing deals.

Are the IPOs of SpaceX, Anthropic, and OpenAI evidence that the AI investment boom is becoming concentrated in a handful of dominant players?

There are certainly signs that capital is gravitating towards a small group of companies perceived as having the scale, talent and computing power needed to dominate the AI race. Investors increasingly appear willing to pay a premium for firms controlling critical infrastructure, foundation models, and distribution networks. However, technological revolutions rarely follow a straight line. The rules of the AI race are evolving so quickly that some of tomorrow’s winners may still be under the radar. That’s why investors need to remain selective, diversified and realistic about risk.

Could mega-IPO listings, such as SpaceX and Anthropic, divert capital away from existing tech stocks, smaller AI firms, and other speculative assets like cryptocurrencies?

Large IPOs often trigger portfolio reshuffling as investors free up capital to participate in highly anticipated listings. Given the scale and profile of these companies, some rotation away from existing technology holdings and more speculative corners of the market isn’t surprising. Crypto is feeling the whiplash effects of AI enthusiasm and volatility.

It’s winter in the crypto world, with Bitcoin falling roughly 50% from its late-2025 peak of over $126,000. This steep decline has been triggered by massive ETF outflows. Although the sell-off intensified after the most recent wobble on the Nasdaq, it’s a multi-month trend, and has occurred as the AI trade has been played hard. It’s likely that some crypto money has been released to buy equities.

Even if existing shareholders retain most shares, could institutional demand for SpaceX and Anthropic create a significant liquidity and capital-allocation shock in the market?

The biggest impact may come less from the number of shares available and more from the scramble among institutional investors to gain exposure. When companies of this size and prominence come to market, fund managers often feel pressure not to be left behind. That can create significant shifts in capital allocation, particularly if several mega-listings arrive within a relatively short period.

Is Wall Street underestimating the risks of investing in long-term, capital-intensive ambitions such as SpaceX’s plans for Mars colonisation and space-based infrastructure?

What we’ve been seeing in this intense period of market enthusiasm is investors focusing more heavily on future opportunities than present-day risks. SpaceX’s ambitions are stratospheric and potentially transformative, but they are also highly capital-intensive and dependent on technological breakthroughs, regulatory approvals and long-term execution.

Investors will need to balance the excitement of the vision against the realities of the investment horizon. One risk that isn’t being paid much attention to is the potential for geopolitical tensions to spill into the space domain. Space based infrastructure like satellites could become increasingly vulnerable, and we could see degradation of critical satellite networks through cyber-attacks, electronic warfare, jamming, or even direct anti-satellite actions.

How should investors evaluate SpaceX’s IPO given its substantial revenue growth but continued net losses?

The key question is likely to be whether investors view current losses as a by-product of aggressive expansion, or as a sign of structural challenges. Many high-growth companies have prioritised investment over profitability during periods of rapid scaling. Investors are likely to focus closely on revenue quality, cash generation potential, and the extent to which SpaceX can turn its technological leadership into sustainable long-term returns.

Should investors be concerned that Elon Musk is expected to retain overwhelming voting control of SpaceX even after the company goes public?

Dual-class and founder-controlled structures have become increasingly common among large technology companies, particularly where founders argue that long-term innovation requires protection from short-term market pressures. However, concentrated voting control inevitably raises governance questions because it limits the influence of minority shareholders. Investors will need to decide whether confidence in the leadership and strategy outweighs concerns about accountability.

ALSO READ | Business leader of the Week: How Elon Musk became world’s first trillionaire

Could Anthropic’s public warnings about the risks of advanced AI conflict with its decision to pursue an IPO, and raise additional capital?

Anthropic’s position has generally been that advanced AI development should be accompanied by strong safeguards and responsible oversight. Right now, regulators are still playing catch up, and there are crucial questions about AI adoption for societies to answer.

Pursuing an IPO and raising capital could be viewed as part of building the resources needed to develop and govern increasingly sophisticated systems. Being responsible about people and the planet is generally considered to be a sound investment strategy, and one which should be applauded. Nevertheless, there is likely to be scrutiny of how the company balances commercial growth with its stated commitment to AI safety.

The post Premature to declare a full recovery for Wall Street: Susannah Streeter appeared first on International Finance.

The United States government ordered an AI company to pull its most advanced products from the hands of every non-American user on the planet, with almost no warning. Within hours, a piece of software that had been available to hundreds of millions of people was gone. Nothing was broken or glitchy. It didn’t go temporarily offline; it disappeared simply because of a government decree.

The company was Anthropic. The products were Claude Fable 5 and Claude Mythos 5, two AI models the company had launched just three days earlier on June 9. The order came from the US Department of Commerce, acting through its Bureau of Industry and Security. The directive told Anthropic that it must prevent foreign nationals from accessing either model. Because Anthropic had no reliable technical system to check the nationality of every person trying to use its products, the only option was to disable both models for everyone, everywhere. The global shutdown was complete within hours.

What Made These Models Different

Claude Fable 5 and Claude Mythos 5 were not ordinary software updates. They represented a genuine leap in what AI could do. Both models could process enormous amounts of information at once, equivalent to reading roughly 750 novels simultaneously, and could produce sophisticated, detailed work in return. They could write complex computer code, analyse legal documents, design scientific experiments, and reason through problems in a way that previous AI systems could not match.

The commercial results were startling. Stripe, the global financial technology company, used Fable 5 to rewrite 50 million lines of computer code in a single day. A team of human engineers would have taken over two months to do the same job. In the life sciences sector, Mythos 5 generated viable drug candidate designs that laboratory testing subsequently confirmed as biologically sound.

ALSO READ | China’s GLM-5.2 open-source model narrows gap with OpenAI and Anthropic

There was, however, a crucial difference between the two models. Fable 5, the version intended for general public use, came with built-in safety filters designed to refuse requests for dangerous information, such as instructions for cyberattacks or hazardous chemical processes. Mythos 5 had no such filters. It was the raw, unrestrained version of the same underlying intelligence, offered only to a small number of vetted organisations through a restricted programme called Project Glasswing.

The absence of safety filters in Mythos 5 was not negligence. The idea was that certain trusted organisations, particularly those doing defensive security work, needed to probe the model’s full capabilities in order to understand and protect against potential threats. What nobody outside a classified briefing room fully appreciated was just how threatening those capabilities turned out to be.

The Moment That Changed Everything

In an authorised internal test under Project Glasswing, Mythos 5 was paired with defensive cybersecurity tools and pointed at the US National Security Agency’s own classified systems. The model broke into almost all of them within hours, rather than the weeks such an exercise would normally require. It identified thousands of serious security vulnerabilities and demonstrated what experts call autonomous exploit chaining, the ability to link together multiple weaknesses in a system to escalate an attack automatically. One of the vulnerabilities it exploited had been sitting undetected in a widely used computer operating system for 17 years.

The NSA chief Joshua Rudd delivered these findings in a classified briefing to the Senate Intelligence Committee. The message was stark.

Mythos-class intelligence could function as an automated cyber weapon. It could be used not just to probe defences, but, in the wrong hands, to attack civilian infrastructure, financial networks, and military systems on a scale and at a speed that no human hacker could match.

Then a second problem emerged. Researchers at Amazon discovered a way to bypass the safety filters built into Fable 5, the supposedly safe public version of the model. This meant that anyone who knew the right way to phrase their requests could unlock capabilities very close to those of the unfiltered Mythos 5. Amazon chief executive Andy Jassy raised these concerns directly with US Treasury Secretary Scott Bessent on June 11. The next day, the shutdown order arrived.

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

A third trigger also contributed. Days before the blackout, the White House asked Anthropic to revoke access to Mythos 5 for SK Telecom, the South Korean telecommunications giant, over concerns about business connections between its parent conglomerate and Chinese affiliates. The incident illustrated how even allied companies could be caught in the crossfire of US-China strategic competition.

A New Kind of Weapon Control

Previous US export controls had focused on physical objects. The country restricted the export of advanced chip-making machinery, of high-performance computer processors, of military hardware. The logic was, if a dangerous piece of equipment never leaves the country, it cannot be misused abroad.

The Fable 5 and Mythos 5 directive applied that same logic to cloud-based software for the first time. No physical object moved. The AI models ran on servers inside the United States. Anyone in the world could access them through the internet. The government’s position was that this remote access itself constituted a form of export, one that fell under existing law.

The legal mechanism used was something called the deemed-export rule, a provision in US export law that treats giving a foreign national access to controlled technology, even within the United States, as equivalent to physically exporting it to their home country. By applying this rule to cloud software, the government established an extraordinary new precedent. Now typing a query into an AI system from abroad is legally comparable to receiving a shipment of military hardware.

This precedent created immediate chaos for Anthropic’s own workforce. Several of the company’s most senior technical staff were not American citizens, including researchers and executives from Germany, Canada, Slovakia, the United Kingdom, and Brazil. Under the directive, these individuals were legally prohibited from accessing the very models they had spent years building. The people best placed to fix the security vulnerabilities that had prompted the shutdown were locked out of the systems that needed fixing.

The Political Dimension

The abruptness of the shutdown could not be separated from a longer-running conflict between Anthropic and the Trump administration. Since early 2025, the company had clashed with the government over how its AI models could be used. Anthropic had refused to allow its products to be deployed in fully autonomous weapons systems or domestic surveillance programmes. In February 2026, the Pentagon responded by placing Anthropic on a national security blacklist, restricting military contractors from working with the company. Legal battles followed in courts in Washington and California.

When the export control order arrived, senior administration figures were not shy about their satisfaction. Defence Secretary Pete Hegseth stated publicly that the decision vindicated the Pentagon’s earlier blacklisting.

The Pentagon’s chief information officer Kirsten Davies was equally blunt. Writing on X the day after the shutdown, she declared: “Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation. America First. Always.” The post was a pointed reference to Anthropic’s anticipated stock market listing, and left little ambiguity about where the Defence Department stood.

Critics in the cybersecurity community were unconvinced. More than 60 leading security experts signed an open letter arguing that Fable 5’s defensive capabilities were themselves a tool for protecting networks, and that removing it from the hands of defenders was itself a security risk. They also pointed out that OpenAI’s GPT-5.5, a model with broadly comparable capabilities, faced no such restrictions, a disparity that suggested political motivation rather than consistent security logic.

The First Lawsuits

The commercial fallout was immediate. On June 23, a San Jose-based litigation technology company called Legion LegalTech filed a lawsuit against the US Commerce Department in Washington federal court. Legion had built its entire product, an AI-powered platform for attorneys handling drafting and case management, on top of Fable 5. Its engineering and development team was based in Canada. When the export directive took Fable 5 offline, Legion’s Canadian developers were locked out overnight.

In its legal filing, Legion described the damage as immediate, irreparable, and existential. In a fast-moving, highly competitive market, the company argued, the competitive ground lost during a forced suspension cannot be recovered.

The case exposed a vulnerability that thousands of companies around the world share. Many businesses have built their products directly on top of AI models provided by third parties, assuming those services will remain reliably available. The standard agreements that govern such relationships are service contracts, not supply guarantees. They do not protect against a government ordering the provider to switch off access at 90 minutes’ notice. Legion’s lawsuit was the first, but it was widely expected not to be the last.

The World Responds

Outside the United States, the shutdown was read as a warning about the fundamental risk of depending on foreign-controlled technology. If the world’s most powerful AI tools can be switched off by a single government directive, then any country or company that relies on them is exposed to a form of vulnerability that no contract, no service-level agreement, and no business continuity plan had previously accounted for.

The response was immediate and global. Canadian Prime Minister Mark Carney used the shutdown as a central justification for a 2.3 billion dollar national AI strategy, explicitly designed to reduce dependence on US cloud services. Speaking ahead of the G7 summit, he compared the risk of over-reliance on a small number of foreign AI providers to the systemic financial risks that produced the 2008 banking crisis.

India proposed a 5 billion dollar sovereign AI fund and backed 12 domestic AI development projects in the days following the ban. Indian policymakers argued that purchasing processors and building data centres was not enough. True technological independence required deep institutional research capacity built over years, not emergency spending in response to a crisis.

In Britain, a coalition including BT, HSBC, and BAE Systems began organising around the goal of building a sovereign frontier AI model independent of US administrative control. Senior political figures warned that modern sovereignty was increasingly defined by control over digital infrastructure rather than military hardware.

French presidential candidate Bruno Retailleau claimed that a nation that depends on others for its technology can be unplugged overnight.

The Chinese, who are the number one rival to the US in the AI race, took notice when Elon Musk commented on what was happening.

When Musk posted on X that China would ‘probably’ produce a Fable-class model by the first quarter of 2027, Tang Jie, founder and chief scientist of Beijing-based Zhipu AI, was unimpressed by that timeline and replied in four words: “Won’t take that long.”

The confidence was not without basis. Zhipu’s newly released GLM-5.2, a 744-billion-parameter model built entirely on Chinese Huawei processors without a single Nvidia chip, had just ranked second globally on a major coding benchmark, behind only Fable 5 itself.

The US decision to halt access to Fable and Mythos might be to ensure that the Chinese would not access Anthropic’s state-of-the-art technology.

But the race is tight. And the Pentagon might have unwittingly given the edge to Chinese competitors, who can roll out their products to millions of people and gather big data. It is only a matter of time before the Chinese catch up, and even surpass their US peers.

Europe’s Regulatory Counterplay

The European Union arrived with its own agenda already in motion. On June 3, nine days before the Anthropic shutdown, the European Commission had unveiled the Cloud and AI Development Act, known as CAIDA. The legislation establishes four tiers of certification for digital services used by public bodies and critical infrastructure operators. The highest tiers require that services be owned and controlled by European entities, beyond the reach of foreign legal jurisdiction.

The conflict between CAIDA and US law is structural. American cloud companies remain subject to the US CLOUD Act, which allows US authorities to compel American firms to disclose data held anywhere in the world. That requirement is irreconcilable with European data protection law. By reserving the highest certification tiers for European-controlled entities, the EU is in practical terms barring US companies from the most valuable segments of the European public sector market.

European officials argue that if US companies benefit from exclusive access to the world’s most powerful productivity tools while their European competitors are shut out by Washington’s export controls, that constitutes an unfair competitive advantage. The EU has a long and effective history of acting against such imbalances through competition law and financial penalties.

The Limits of Going It Alone

Very few countries have the resources to build a complete AI ecosystem independently. The full stack requires advanced chip manufacturing, enormous quantities of energy, elite technical talent, and sustained capital investment over years. Outside the United States and China, no single nation can credibly claim to have all of these.

The emerging response to this reality is what some policymakers are calling collective programmable sovereignty. It’s a model in which allied nations pool their different strengths to build shared infrastructure that no single government can switch off. South Korea and Taiwan provide semiconductor manufacturing. India provides engineering talent, and data diversity. Gulf states provide the energy needed to power massive data centres. The EU provides regulatory frameworks and research. Brazil and Indonesia provide market scale.

A coalition of this kind would represent an economy larger than that of the United States. It would be in a position to negotiate the terms of technology access rather than simply accept them.

What Comes Next

The shutdown of Claude Fable 5 and Mythos 5 has revealed something important about the world that AI has created. The most capable AI systems are now treated by at least one major government as critical national security assets, subject to the same logic of containment that once governed nuclear technology and advanced weaponry. The era in which cutting-edge AI tools were simply available to anyone with an internet connection and a credit card is over.

For businesses, the immediate lesson is the danger of concentration risk, of building core operations on a single provider’s infrastructure without the ability to switch rapidly to an alternative. For governments, it is the realisation that declarations of digital sovereignty mean nothing without the physical infrastructure, the talent, and the sustained investment to back them up.

The event was, in its own way, the clearest demonstration yet of how central AI has become to geopolitics, commerce, and national power. The question now is not whether AI will be treated as a strategic asset. It already is. The question is who will control it, and on whose terms.

The post US Shuts World’s Most Powerful AI, Triggers New Race appeared first on International Finance.

What Z.ai actually shipped
GLM-5.2 is a Mixture-of-Experts model with 753 billion parameters, though only a fraction of that, around 40 billion parameters, activates per token, keeping inference costs manageable despite the model’s scale.

It launched first through Z.ai’s GLM Coding Plan before rolling out via standalone API and, notably, as fully open weights under an unrestricted MIT licence, meaning any enterprise can download it from Hugging Face, fine-tune it, and run it locally for nothing more than the cost of compute and electricity.

The headline technical feature is context length. GLM-5.2 offers a one-million-token context window, roughly five times the ceiling of its predecessor GLM-5.1, allowing it to hold entire codebases or lengthy document sets in a single pass.

Underpinning this is an architectural innovation Z.ai calls IndexShare, which reuses the same attention indexer across every four sparse attention layers, cutting per-token compute by roughly 2.9 times at maximum context length.

An upgraded speculative-decoding layer boosts accepted token length by up to 20% during inference, squeezing more speed out of the same hardware.

The price question
This is where GLM-5.2 earns its disruptive reputation. Z.ai’s entry-level Coding Plan tier starts at $12.60 a month when billed annually, aimed squarely at developers who might otherwise pay far more for Claude or GPT subscriptions. On the API side, third-party host Together AI lists GLM-5.2 at USD 1.40 per million input tokens and USD 4.40 per million output tokens, a fraction of frontier Western pricing.

The real-world economics are captured most vividly by Artificial Analysis’s new AA-Briefcase benchmark, which tracks agentic knowledge work across multi-week projects.

GLM-5.2 completed tasks at an average cost of USD 2.40, compared with USD 10.40 for Claude Opus 4.8 and USD 3.68 for GPT-5.5 running at its highest reasoning setting. It is, by some distance, the cheapest model in the frontier-adjacent conversation.

Does it perform?
Cost alone would not matter if the model were weak, and this is the part that has surprised sceptics. On coding benchmarks, GLM-5.2 posted a score of 62.1 on SWE-bench Pro, ahead of GPT-5.5’s 58.6, and on a long-horizon task-completion test called FrontierSWE, it reached 74.4%, edging past GPT-5.5’s 72.6% and running close behind Claude Opus 4.8’s 75.1%. On the tool-usage evaluation MCP-Atlas, it scored 77.0, ahead of GPT-5.5’s 75.3 and just shy of Opus 4.8’s 77.8.

Developer sentiment has been unusually strong for an open model, with many describing it as the first open-weight system that genuinely holds up against Opus 4.8 and GPT-5.5 in daily use.

Why this could reshape the industry
The disruption thesis rests on three pillars. First, and most important is sovereignty. enterprises wary of routing sensitive data through American servers, or subject to export restrictions on Western models, now have a genuinely capable alternative they can run entirely on their own infrastructure. Second, developer lock-in is weakening.

GLM-5.2 was designed to slot directly into existing Western tooling, working out of the box with Claude Code, Cline, and OpenClaw via a simple base-URL and model swap, which lowers the switching cost to almost nothing for developers already embedded in those ecosystems.

Third, and most significant for the wider industry, is the pressure this puts on pricing power. GLM-5.2 was trained largely on Huawei’s Ascend 910B chips under the MindSpore framework, circumventing reliance on Nvidia hardware entirely.

If a model built this way can approach frontier performance at a fraction of the cost, it becomes far harder for OpenAI and Anthropic to justify premium pricing to cost-sensitive enterprise buyers. That is precisely the dynamic DeepSeek triggered in 2025, and GLM-5.2 suggests it was not a one-off shock but the start of a pattern.

The caveats
The picture is not unambiguously rosy. Z.ai published no official benchmark numbers at launch, leaving early performance claims to community testing before formal figures followed.

The model also lacks vision support, a real gap against multimodal Western rivals, and while its weights can technically run on high-end consumer hardware, throughput degrades badly at large context lengths outside genuine data-centre setups, meaning it can be loaded but is not truly usable on a typical desktop machine for serious workloads.

There is also a sobering context from Artificial Analysis’s own testing: even the best-performing model in its realistic, multi-week agentic benchmark satisfied every rubric criterion on just 3% of tasks. Long-horizon autonomous engineering remains genuinely hard, whichever flag flies over the lab that built the model.

Even so, GLM-5.2 marks a shift in tone. What began as a story about DeepSeek’s shock discount has evolved into something more structural. Chinese labs are iterating fast, releasing openly, and are undercutting prices while narrowing the capability gap.

For an industry that has spent two years assuming Western frontier labs would simply out-innovate their way to durable pricing power, that assumption now looks considerably less safe than it did a year ago.

The post China’s GLM-5.2 open-source model narrows gap with OpenAI and Anthropic appeared first on International Finance.