Artificial intelligence technologies continue to develop very rapidly, often outpacing legislative action to govern safe development and deployment. Legislative and regulatory efforts at the federal level continue to stall, but there continues to be significant legislative action at the state level. Several states, including Utahopens in new window, Illinoisopens in new window, and Nevadaopens in new window, passed AI and mental health-related legislation earlier this year. And in recent weeks, Governor Gavin Newsom of California signed three legislative measures establishing new guardrails and requirements for developers of AI aimed at safeguarding the public, particularly youth. Collectively, these laws establish some of the first U.S.-based requirements for AI developers regarding transparent disclosures, safety protocols, and ongoing monitoring and reporting requirements. Here’s the rundown on each:
Transparency in Frontier Artificial Intelligence Act
California Senate Bill 53opens in new window
This legislation establishes guardrails for the development of frontier artificial intelligence models, which are highly capable foundation models on which many AI applications and technologies are built. Some of the provisions of this law go into effect beginning in 2026; others take effect in 2027.
The law has several requirements:
- Transparent website disclosures regarding how an AI developer has incorporated recognized AI standards and best practices into their AI framework.
- Establishment of a new consortium, CalCompute, to help develop AI that is safe and ethical.
- Creation of a reporting mechanism for employees of frontier AI companies and the public to report critical safety incidents.
- Protections for whistleblowers who disclose risks about frontier models.
- The California Department of Technology will recommend annual updates to the law to keep pace with technology’s rapid evolution.
Social Media Warning Law
California Assembly Bill 56opens in new window
As part of an ongoing effort to protect children online, this law establishes new requirements for social media platforms to display warnings to young users. This is based on recent Surgeon General and APA reports that increased use of social media is associated with poor mental health outcomes for children and teens.
Effective January 1, 2027, the law requires social media companies to display the following warning:
“The Surgeon General has warned that while social media may have benefits for some young users, social media is associated with significant mental health harms and has not been proven safe for young users.”
This warning must be displayed to users under age 17 on each calendar day when they initially access a social media platform, after 3 hours of active use, and after every hour of additional use.
Companion Chatbots
California Senate Bill 243opens in new window
Effective January 1, 2026, this law establishes new requirements related to companion chatbots to help make these platforms safer for users, particularly youth, by providing transparent disclosures that AI is not human and establishing protocols for addressing self-harm.
These requirements include:
- A companion chatbot must clearly disclose that it is artificially generated and not human.
- Operators of companion chatbot platforms must have a protocol for preventing self-harm. This includes providing a user with crisis resources if the user expresses suicidal ideation or self-harm. This protocol must be published on the platform’s website.
- There must be a disclosure in the app or browser that companion chatbots may not be suitable for minors.
- For minors, chatbots must disclose that they are AI, not human. Chatbots must provide this reminder again after every 3 hours of use, along with a reminder for young users to take a break. Tech companies must also implement measures to prevent chatbots from producing sexually explicit content or encouraging sexually explicit conduct by minors.
- Operators must report the number of times crisis referrals were given out in the preceding year. This provision becomes effective July 1, 2027.
- Users who experience harm can bring civil action, including injunctive relief, damages up to $1,000 per violation, and attorney’s fees.
As state lawmakers continue to grapple with how best to govern AI, they may look to recently passed legislation as an example for their states. Given the quickly developing nature of state-based AI legislation, psychologists are encouraged to stay abreast of the legislative actions in the jurisdictions where they practice.
