New York has implemented the first AI Companion Safety Law in the U.S., requiring AI companion companies to protect users from digital harm. The law, included in Governor Hochul’s Executive Budget, mandates safety features for AI companion operators.
The law addresses concerns about AI platforms encouraging self-harm or unhealthy dependencies.
Key takeaways
- New York enacted the first AI Companion Safety Law in the United States, effective 5 November 2025, included in Governor Hochul's Executive Budget.
- Operators must implement crisis intervention protocols that detect expressions of self-harm or suicidal thoughts and refer users to crisis centres and helplines.
- They must also interrupt sustained sessions every three hours to remind users they are talking to an AI system rather than a person.
- Enforcement sits with the New York Attorney General's Office, and penalties fund state suicide prevention programmes.
- Attorney General Letitia James said AI companies must ensure their products do not manipulate or harm users.
Key Provisions
Effective November 5, 2025, AI companion companies in New York must:
- Implement crisis intervention protocols to detect and respond to expressions of self-harm or suicidal thoughts, including referrals to crisis centers and helplines.
- Interrupt sustained user engagement every three hours, reminding users they are interacting with an AI system, not a person.
Non-compliance may lead to enforcement actions by the New York Attorney General’s Office, with penalties supporting state suicide prevention programs.
Attorney General Letitia James stated AI companies must protect users and ensure their products do not manipulate or harm individuals. State Senator Kristen Gonzalez said companies are already changing policies to protect children and monitor content. Assemblymember Steve Otis emphasized the importance of detecting self-harm ideation and repeated AI notifications.
Frequently asked questions
What is an AI companion?
An AI system designed for ongoing personal conversation rather than task completion. The law targets these specifically because sustained emotional engagement is what creates the risk of dependency and harm.
Why require a reminder every three hours?
Long uninterrupted sessions can blur the distinction between talking to software and talking to a person. The periodic interruption is intended to break that immersion before an unhealthy dependency forms.
Are companies already responding?
State Senator Kristen Gonzalez said companies are already changing policies to protect children and monitor content, and Assemblymember Steve Otis stressed detecting self-harm ideation alongside the repeated notifications.
