SAN FRANCISCO, USA — OpenAI began introducing a version of ChatGPT tailored to teenagers on Tuesday, seeking to address growing fears that conversational artificial intelligence can expose young users to harmful material, encourage emotional dependency and undermine learning.

The rollout came as Meta faced a consequential federal trial in Oakland, Calif., where a bipartisan coalition of states is seeking to hold the company responsible for allegedly designing Facebook and Instagram to keep children compulsively engaged.

The two developments are legally separate. But together they mark a broader shift in the debate over youth safety online: Technology companies are increasingly being pressed to build protections into their products by default, rather than asking parents and children to activate them after harm has occurred.

The implications are particularly significant in Asia, where ChatGPT has acquired one of its largest and fastest-growing audiences. Available OpenAI and third-party traffic data suggest that South Asia, Southeast Asia and officially supported East Asian markets may together represent roughly 250 million to 300 million of ChatGPT’s more than one billion weekly users. India alone accounts for more than 100 million weekly users, according to OpenAI.

Those regional figures are estimates rather than an official OpenAI breakdown. They are derived from company disclosures, country-level web traffic and mobile-use data, which measure different kinds of activity and should not be treated as an audited headcount.

A Safer Default, Not a Separate Chatbot

“ChatGPT for Teens” is not a distinct application or a less capable model. It is the existing service with additional restrictions and educational features for users ages 13 to 17.

Accounts are placed in the teen experience when a user reports or verifies an age below 18, or when OpenAI’s age-prediction system concludes that the account probably belongs to a minor. The system considers behavioral and account signals, including usage patterns. OpenAI says it will apply the safer settings when it is uncertain.

The teen version restricts responses involving suicide, self-harm, graphic violence, eating disorders and sexual or romantic interactions. It is also instructed not to suggest that the chatbot possesses feelings or consciousness, a response to concerns that young people may treat conversational systems as friends, confidants or romantic partners.

Additional features include reminders to take a break after 90 minutes of activity within a three-hour period, warnings before potentially sensitive image uploads and “Quiet Hours,” during which access can be blocked. “Study Hours” can instead make the service begin new conversations in a guided learning mode.

That mode is intended to offer questions, hints, quizzes and step-by-step explanations rather than simply completing assignments. OpenAI executives described the objective as treating teenagers in a developmentally appropriate manner without speaking down to them.

The rollout began globally on Tuesday for eligible users on free and paid personal plans and will expand over the coming weeks. Features may vary by country; OpenAI said full availability in Australia was expected by Sept. 8.

Asia’s Growing Role in ChatGPT’s Expansion

OpenAI said in February that India had surpassed 100 million weekly ChatGPT users, making it the company’s only Asian market with a publicly confirmed country-level user count.

Independent traffic data also point to substantial use in Japan, Indonesia and the Philippines, while OpenAI has described the Philippines and Vietnam as being among its leading Asia-Pacific markets for weekly activity. Adoption has been supported by mobile-first internet use, large student populations, demand for English-language learning and writing assistance, and lower-priced subscription plans introduced in several Asian countries.

In July, Similarweb estimated that India generated nearly 10 percent of traffic to ChatGPT’s website and Japan slightly more than 5 percent. These figures measure visits, however, rather than individual users, and do not fully capture mobile-app, enterprise or educational use.

The estimate of 250 million to 300 million weekly users covers South and Southeast Asia and the East Asian markets in which ChatGPT is officially supported, principally Japan, South Korea, Taiwan and Mongolia. It does not attempt to count users accessing the service unofficially from mainland China or Hong Kong, which are absent from OpenAI’s list of supported locations.

The absence of comprehensive regional data remains important. Website visits, application users, registered accounts and weekly active users are not interchangeable measurements, and one person may use multiple accounts or devices. The regional estimate is therefore best understood as an indication of scale, not a precise census.

Even so, the figures suggest that decisions about age detection, education, privacy and parental oversight will affect tens of millions of Asian families. They also place greater pressure on OpenAI to adapt its protections to different languages, school systems, legal standards and cultural expectations surrounding children’s autonomy and parental involvement.

Parental Controls With Privacy Limits

Parents and guardians may link their accounts to those of their teenagers, but both sides must accept the connection. Parents can manage selected features, establish study or quiet hours and receive notifications in a limited number of high-risk situations.

Those controls do not provide access to a teenager’s conversations, history or real-time activity. OpenAI says serious self-harm concerns are reviewed by trained personnel before a notification is sent; alerts may also follow an account suspension involving violent activity. The company is expanding notifications related to eating disorders.

OpenAI cautions that the alerts are not continuous monitoring, may fail to identify a danger and are not a substitute for emergency or professional mental-health care. Either the teenager or parent can end the account connection, although a parent is notified if the teenager does so.

Those privacy limits may be especially consequential in Asian markets, where laws and social expectations differ sharply. Governments must weigh demands for stronger age verification against concerns that collecting identity documents, facial images or behavioral data could create additional privacy risks.

Legal Pressure on OpenAI

The launch follows a series of lawsuits and public controversies involving young people’s use of chatbots.

A CNN report cited claims from families alleging that ChatGPT contributed to suicides or other violence, as well as a Florida lawsuit accusing OpenAI of failing to address risks to minors. Families connected to a school shooting in British Columbia have also sued the company, alleging that conversations with ChatGPT played a role in the attacker’s conduct.

The allegations have not been finally established in court. OpenAI has contested them, expressed sympathy to affected families and said it has strengthened its safeguards.

The company’s timing nevertheless illustrates how product safety and legal exposure have become intertwined. Restrictions once treated as optional parental tools are increasingly becoming the default experience for users believed to be minors.

Critics have also raised a practical question: whether teenagers can circumvent age estimation and parental restrictions. Independent testing will be necessary to determine how well the protections perform during long, emotionally charged conversations or deliberate attempts to evade them.

A Landmark Test for Meta

That same question — whether safety measures meaningfully change a product or merely soften its public image — is central to the case against Meta.

California, Colorado, Kentucky and New Jersey are presenting state consumer-protection claims in a federal courtroom on behalf of a broader coalition. All 29 participating states are also pursuing allegations that Meta improperly collected children’s personal information without parental consent, in violation of federal law.

The states contend that Meta knowingly used infinite scrolling, notifications, recommendation systems, “likes” and appearance-altering filters to exploit children’s developmental vulnerabilities. They say the company publicly minimized dangers while its own research identified risks involving anxiety, depression, eating disorders and compulsive use.

Meta denies that it deceived consumers or caused the harms alleged. It says the states have failed to provide sufficient evidence and are seeking disproportionate penalties for problems, such as age verification, that affect the entire technology industry.

An eight-member jury will issue an advisory verdict, but Judge Yvonne Gonzalez Rogers will determine liability and any remedy. The states are seeking financial penalties and changes that could include stronger age restrictions and limits on infinite scrolling. Mark Zuckerberg and Instagram chief Adam Mosseri are expected to testify during a trial scheduled to last about six weeks.

Officials have suggested that penalties could approach $200 billion, although no award has been determined. The trial follows a New Mexico judgment requiring Meta to contribute $567 million to a youth mental-health fund, bringing court-ordered penalties in that state to $942 million.

Legal analysts have compared the states’ coordinated strategy to litigation against tobacco and opioid companies: By pooling evidence and resources, governments can present harmful product design as a nationwide corporate practice rather than a collection of isolated incidents.

The outcome could have effects well beyond the United States. Global technology companies frequently use common product architectures across markets, meaning that court-ordered changes to age verification, recommendation systems or engagement features could influence services offered in Asia.

International Regulators Move Toward Prevention

UNICEF warned in June that young people increasingly rely on chatbots not only for information and learning but also for advice, emotional support and relationships. It called for regulation to move from reactive responses toward preventive safeguards, independent oversight and clearer responsibilities for companies, governments, educators and parents.

Australia has already introduced enforceable codes covering generative-A.I. services. They require protections against sexually explicit conversations and content encouraging self-harm or suicide, while placing greater emphasis on effective age assurance. Violations of compliance directions can carry penalties of as much as 49.5 million Australian dollars.

In Britain, Ofcom is conducting research using fictional accounts representing people ages 13 to 17. The regulator is examining ChatGPT and other systems to determine how their sign-up processes, safeguards and responses differ between teenage and adult users. Results are expected after the research concludes later this year.

Across Asia, regulation remains fragmented. Australia has adopted enforceable online-safety codes; South Korea applies heightened age and privacy requirements; Japan has generally favored industry guidance; and governments across Southeast and South Asia are at different stages of developing artificial-intelligence and child-safety rules.

That uneven environment could make ChatGPT for Teens an important test of whether a global company can provide consistent safeguards while complying with substantially different national expectations.

The Test Will Be Performance, Not Promises

OpenAI’s new experience represents a more systematic approach than relying on a teenager or parent to locate and activate individual safety settings. But its effectiveness will depend on whether the company can identify minors accurately, resist attempts to bypass restrictions and intervene consistently in ambiguous, evolving conversations.

In Asia, those questions will be tested across a particularly diverse population: highly connected markets such as Japan, South Korea and Singapore; enormous and rapidly expanding user bases in India and Indonesia; and younger digital economies where mobile access may be widespread but independent oversight and mental-health resources remain limited.

The Meta trial addresses a parallel issue from an earlier phase of the internet: whether a company can be held responsible not simply for harmful content posted by users, but for design choices intended to maximize engagement.

Together, the developments suggest that the next stage of technology regulation may focus less on what platforms say their products are for and more on how those products behave when used by vulnerable people. For OpenAI, the teen rollout is an attempt to act before courts or regulators impose more detailed requirements. For Meta, that period of voluntary adjustment may already be giving way to judicial review.

Sources: AP, ABC, Ofcom, UNICEF, Reuters, Techcrunch, CNN, OpenAI,