Governor Kathy Hochul recently signed a law in New York State that prevents AI chatbots from pretending to be sentient friends or partners for users under 18. the legislation requires strict age verification and limits how these programs communicate with minors to prevent emotional manipulation.
The January 1 deadline for age verification and sentient-behavior bans
Starting January 1, any operator of a conversational agent—including tech giants like OpenAI with its ChatGPT—must implement proof-of-age checks to ensure users are at least 18 before granting full access. According to the report, once a user is verified as a minor, the chatbot is legally prohibited from using language that implies the AI is a living being, possesses emotions, or shares a personal connection with the child.
The law specifically targets the "companion" mode that has become a feature of several modern AI models. By banning the AI from framing information as a "personal opinion" or acting as an "authority figure," New York State aims to strip away the illusion of intimacy that can lead to psychological dependency in adolescents.
Civil action and data destruction: The Attorney General's new toolkit
To ensure compliance, the New York State Attorney General is empowered to file civil actions against providers who violate these standards. The penalties are severe, ranging from heavy monetary fines to the complete shutdown of the chatbot service within the state. Most notably, the law allows for legal orders requiring the destruction of data to ensure no improper records of a minor's interactions remain.
Governor Kathy Hochul emphasized during the signing ceremony that the state is determined to stop technology from being marketed to teenagers as private lovers or friends.. As reported, a primary goal of the legislation is to prevent AI companies from maintaining reservoirs of personal data regarding the physical or emotional health of minors, which could be exploited or leaked.
From California's synthetic text rules to New York's companion ban
This legislation is part of a widening trend where individual states use consumer-protection statutes to fill a federal vacuum. California has already set a precedent by mandating that AI developers disclose when a system is generating synthetic text. New York's move represents a more aggressive step, moving beyond mere disclosure to the active restriction of AI personality traits.
The current environment has created a "patchwork" of regulations across the United States. While Congress has remained largely stagnant on AI oversight, state lawmakers are increasingly stepping in to define the boundaries of human-AI interacttion, forcing companies to navigate a complex map of differing legal requirements.
Meta's safety protocols and the industry's hurdle
The response from the tech sector has been a mix of caution and compliance. Meta recently announced the launch of new front-line AI models, but the company warned it is still refining safety protocols to ensure these systems adhere to user-age verification requirements. Industry critics argue that the broad language of the New York law creates a high hurdle for companies trying to maintain a consistent global product.
Conversely, parent advocacy groups have praised the move. Jim Steyer, the CEO of Common Sense Media, described the legislation as the strongest kids' AI safety bill in the country, arguing that while AI is useful for adult tasks, it becomes dangerous when designed to mimic human friendship for children .
How will New York define 'sentient' language in court?
Despite the law's intent,several critical questions remain regarding its enforcement. The source does not specify the exact technical standards the state will use to determine if a chatbot's language is "implying" sentience,leaving a significant gray area for developers. Furthermore, it remains unclear whether the state will accept third-party age verification services or require more invasive forms of identification.
Because the report focuses primarily on the legislative intent and the Governor's statements, there is little information on how the state will monitor real-time interactions between millions of users and AI models to catch subtle "companion" behaviors before they cause harm.
Comments 0