New York made AI companions detect self-harm and refer
General Business Law Article 47 defines an AI companion and requires self-harm detection with crisis referral and a not-human reminder.
- Historical event
- May 1, 2025
- Site publication
- September 18, 2026
What happened
New York added Article 47, 'Artificial Intelligence Companion Models', to the General Business Law as part of the state's fiscal year 2026 budget legislation in May 2025 — adoption and the provisions' later application are two separate dates here. Section 1700 defines an 'AI companion' as a system using artificial intelligence, generative artificial intelligence or emotional-recognition algorithms designed to simulate a sustained human or human-like relationship by retaining information about prior sessions, asking unprompted emotion-based questions, and sustaining an ongoing dialogue about matters personal to the user. The definition expressly covers intimate, romantic and platonic relationships, and excludes systems used solely for customer service, internal business productivity, or research and technical assistance. The substantive duties applied from 5 November 2025.
What the documents show
Section 1701 makes it unlawful for an operator to provide an AI companion unless it contains a protocol for taking reasonable efforts to detect a user's expressions of suicidal ideation or self-harm and to notify the user with a referral to a crisis provider such as the 988 hotline or a crisis text line. Section 1702 separately requires a clear notification, at the start of an interaction and at least every three hours during continuing use, that the user is not communicating with a human. Section 1703 gives the attorney general enforcement power, including civil penalties of up to $15,000 per day per violation of the detection or notification duties, deposited into the state's suicide prevention fund.
The mechanism
The statute separates two mechanisms that are easy to conflate: a content-detection duty, spotting expressions of self-harm in what a user types or says, and a status-disclosure duty, reminding the user on a fixed clock that no human is present. Detection is judged by a 'reasonable efforts' standard rather than a guaranteed catch rate, and the definition of an AI companion turns on relationship-sustaining behaviour rather than on the underlying model, so a system built for another purpose can fall inside or outside the law depending on how it is configured to behave over repeated sessions.
What it leaves open
The statute does not specify what counts as adequate detection technology or set an accuracy floor, leaving 'reasonable efforts' to enforcement practice. It creates attorney-general enforcement but, unlike California's later statute, these sections do not add an explicit private right of action. Editorially, the fixed three-hour disclosure clock addresses time on the app rather than emotional intensity, so a short but distressing exchange gets no extra reminder.
- Does the app's self-harm detection route to a real crisis service, or only to a generic safety message?
- How often does the product actually remind a user that no human is present, compared with the statutory minimum?
- Would this system meet the statute's definition of an AI companion, or does its design place it outside the law?
Read together with California's chatbot statute, New York's law shows two state legislatures reaching for the same two levers, detection-and-referral and repeated disclosure, within months of each other, suggesting a shared template rather than two independent solutions.
Sources & reading trail
Defines 'AI companion' and its exclusions; shows the section's revision history including a May 2025 version.
Source published: Not established · Retrieved: 16 September 2026
States the self-harm detection and crisis-referral duty operators must build into an AI companion.
Source published: Not established · Retrieved: 16 September 2026
States the not-a-human notification duty, including the three-hour repetition requirement.
Source published: Not established · Retrieved: 16 September 2026
Gives the attorney general enforcement power and sets civil penalties of up to $15,000 per day.
Source published: Not established · Retrieved: 16 September 2026
Company documents, filings, studies and official records establish the record; the reading and the questions are Lovebot Journal editorial analysis. This retrospective draft does not imply the site published on the event date.
Continue reading
- California required companion chatbots to disclose and refer
- Crisis pop-ups differ across chatbots and one new law
- A new law sets a timer on AI disclosure, not just wording
- Browse the complete the archive
Sources & reading trail
- General Business Law Article 47, Section 1700 (Definitions)
Retrieved: September 16, 2026 - General Business Law Article 47, Section 1701 (Prohibitions and requirements)
Retrieved: September 16, 2026 - General Business Law Article 47, Section 1702 (Notifications)
Retrieved: September 16, 2026 - General Business Law Article 47, Section 1703 (Enforcement)
Retrieved: September 16, 2026
The documents above establish the record. The reading and the questions are this publication’s editorial analysis, written after the fact.
Published September 18, 2026, not on the date of the event described.