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.”