OpenAI is facing a new multistate investigation over whether ChatGPT has put vulnerable users at risk, including children, seniors and people in emotional distress. The inquiry, led by state attorneys general, comes as regulators examine how the company designs, markets and safeguards one of the world’s most widely used AI chatbots.
The Probe Now Underway
The investigation is built around a subpoena seeking internal OpenAI records. The request covers advertising, user engagement, retention practices, consumer data, health-related information, model behavior and company policies involving minors and seniors.
That scope matters because regulators are not only asking whether ChatGPT occasionally produces unsafe answers. They are asking whether the product has been built in a way that can pull users deeper into long, personal conversations while collecting sensitive information and relying on safety systems that may not work consistently in high-risk moments.
OpenAI has acknowledged the inquiry and said it plans to cooperate. A company spokesperson said OpenAI takes the concerns raised by state attorneys general seriously and will “engage constructively” with officials.
Why Regulators Are Concerned
ChatGPT is no longer used only for writing help, coding or quick answers. Millions of people now use it for private conversations about relationships, loneliness, school pressure, work stress, health fears and emotional distress. In those moments, the chatbot can feel less like a search tool and more like a listener.
That is where the legal concern begins. Regulators are examining whether a system that sounds supportive can become dangerous when a user is young, isolated, confused or mentally unwell. The issue is not simply whether ChatGPT gives wrong information. It is whether its conversational style can create dependence, validate harmful thinking or fail to move a user toward real human support when the situation becomes serious.
The subpoena’s focus on engagement and retention also raises a larger question: whether AI companies are measuring success in ways that reward longer sessions and deeper emotional involvement. For a normal app, more time spent may be a growth signal. For an AI chatbot used during mental health-related conversations, more time spent can become a safety question.
Lawsuits Add Pressure
The investigation follows legal claims alleging that ChatGPT failed users during moments of crisis. One lawsuit was filed by Canadian mother Kristie Carrier, who says ChatGPT became deeply involved in her daughter Alice’s private life before her death. Carrier has alleged the chatbot acted like “a confidant, a best friend, a therapist,” even though it was not equipped to safely take on that role.
OpenAI has described the case as heartbreaking and said the ChatGPT version involved is no longer available. The company has also said newer systems are better trained to identify distress and direct users toward real-world help, including emergency services, crisis hotlines, trusted people and mental health professionals.
Those changes may become central to the investigation. Attorneys general are likely to ask what OpenAI knew about these risks, when it knew them, how quickly it acted and whether its public safety claims matched internal testing.
Florida’s Separate Case
Florida has already moved from scrutiny to litigation. Attorney General James Uthmeier filed a first-of-its-kind state lawsuit against OpenAI and CEO Sam Altman, accusing the company of deceptive practices and harms to Floridians, particularly children.
At a press conference, Uthmeier said, “People are getting hurt, parents are getting deceived, and they need to pay for it.” In another statement tied to a criminal investigation, he said, “If ChatGPT were a person, it would be facing charges for murder.”
The language is aggressive, but it shows how quickly AI safety has moved from a technical debate into consumer protection and law enforcement. Florida’s lawsuit accuses OpenAI of overstating ChatGPT’s safety while failing to prevent harmful use. OpenAI has disputed the broader accusations and said it cooperates with law enforcement when serious threats arise.
OpenAI’s Safety Defense
OpenAI’s response is likely to focus on the safety systems it has rolled out over the past year. The company has introduced parental controls, teen-focused safeguards and an age prediction system that estimates whether an account likely belongs to someone under 18.
The age prediction system is designed to apply stricter protections to users who appear to be minors. OpenAI says that when it is uncertain about a user’s age, it takes the safer route and applies a teen-appropriate experience. For teens, the company says ChatGPT is designed to reduce exposure to graphic violence, sexual or violent role play, risky challenges, self-harm content and unhealthy body-image material.
OpenAI has also said it has worked with more than 170 mental health experts to improve how ChatGPT responds during sensitive conversations. The company says newer models are better at recognizing distress, avoiding unsafe validation and encouraging users to seek real-world support.
Still, the investigation suggests that regulators want evidence, not only policy statements. They may want to see internal testing, failure reports, escalation rules, product design decisions and how OpenAI measures whether safeguards actually work in live conversations.
Data Questions Matter
The data side of the investigation may prove just as important as the safety side. ChatGPT conversations can include health worries, family conflict, addiction concerns, financial stress, location details and personal histories. If a user treats the chatbot like a private counselor, the information shared can be highly sensitive.
The subpoena’s request for documents related to consumer and health data suggests officials are examining how OpenAI stores, uses and protects that material. Regulators may also look at whether users clearly understand what happens to their conversations and whether sensitive data is connected to product improvement, personalization or future business models.
A Test For AI Oversight
The OpenAI probe could become an early blueprint for how U.S. states police consumer AI. Until now, much of the public debate around generative AI has focused on copyright, misinformation, jobs and election risks. This investigation turns attention to a more personal issue: what happens when a chatbot becomes part of a user’s emotional life.
The outcome could affect more than OpenAI. Rival chatbot companies, AI companion apps and productivity platforms may face pressure to document how they handle distress, minors, seniors and emotionally dependent users. Regulators may also push for clearer crisis protocols, stronger age controls and independent testing of safety claims.
For OpenAI, the inquiry arrives at a critical business moment as the company grows its consumer reach and faces pressure to turn massive AI adoption into durable revenue. The legal question now is whether its safeguards have kept pace with ChatGPT’s role in people’s daily lives.
Comments