Which AI Bills Were Signed into Law in 2025?
Key highlights this week:
We’re tracking 1,080 bills in all 50 states related to AI during the 2025 legislative session.
The California Privacy Protection Agency board unanimously adopted a package of regulations related to automated decision-making technology. Separately, a federal judge struck down the state’s political deepfake law.
Colorado lawmakers will return for a special session this month to address issues with its landmark AI law set to go into effect in 2026.
The governor of Illinois signed a bill into law to prohibit the use of AI in mental health services without informed consent.
By our count, this year, lawmakers in all 50 states have considered 1,080 bills related to AI. Yet despite this clear interest from lawmakers, only 118 bills have been signed into law this year (less than 11%). This week, we’ll take a look at which bills are making it across the finish line.
Similar to 2024, deepfake legislation has been of major interest to lawmakers this year. Of the 1,080 AI bills introduced, 301 have addressed deepfakes. Of those 301 bills, 68 were signed into law. Furthermore, 68 of the enacted bills addressed sexual deepfakes specifically (as opposed to political deepfakes; however, many states did adopt political deepfake laws last year in the lead-up to the 2024 election). As we have previously discussed, deepfake legislation is an easy target for state lawmakers, particularly where sexual deepfakes are concerned. This has also been true on the federal level, where the only major AI-specific law to be enacted has addressed sexual deepfakes.
This is likely a trend that will continue in the coming years to address other concerns surrounding AI-generated images and videos. For example, a subset of deepfake legislation addressing the use of digital replicas (authorized AI-generated images of individuals for commercial use) has gained interest among state lawmakers, with lawmakers in 22 states considering digital replica legislation. Of these, four states that enacted deepfake legislation: Arkansas (AR HB 1071), Montana (MT HB 513), Pennsylvania (PA SB 649), and Utah (UT SB 271) enacted digital replica legislation this year.
Legislation addressing the use of AI in healthcare, including health insurance, has also had some success getting enacted this year, with 9 of 83 bills being signed into law. Enacted legislation addressing the use of AI in healthcare is quite varied, and includes prohibitions on AI as the sole decision maker in the utilization review process (NE LB 77), prohibiting using AI to practice mental or behavior healthcare (NV AB 406), and requiring human review of AI used for diagnostic purposes (TX SB 1188). This week, the governor signed a bill (IL HB 1806) into law in Illinois to prohibit the use of AI in mental health services without informed consent. However, legislation addressing AI’s use in healthcare has also faced some setbacks, with provisions related to AI ultimately struck in bills in Arizona (AZ HB 2175) and Connecticut (CT SB 10), which removed provisions related to AI’s use in the utilization review process prior to those bills being signed into law.
Another area where states have been successful in enacting AI legislation relates to state uses of AI. This year, 13of 64 introduced bills were enacted in seven states addressing state use of AI. Texas was the most active in this area, enacting legislation establishing an AI Division to assist agencies in implementing AI technology (TX HB 2818), requiring AI training for certain state employees (TX HB 3512), and requiring an inventory of state AI systems (TX SB 1964). Texas also enacted legislation that placed a number of restrictions on state uses of AI (TX HB 149), including requiring disclosures when individuals are interacting with a state AI system, prohibiting AI from infringing on constitutional rights, and prohibiting the use of AI to identify an individual through biometrics without their consent.
This interest in state AI use has not been confined to the 50 states. In Puerto Rico, three state AI bills have moved out of their chamber of origin in the Legislative Assembly (PR PC 347, PR PS 622). This includes a bill (PR PS 68) that would create the position of AI officer in the Puerto Rican government to ensure all government agencies using AI systems are in compliance with applicable laws and regulations.
There is no reason to think that state lawmakers are likely to lose interest in this issue or grow frustrated with the slow progress of bills getting signed into law. More likely, the small percentage of enacted legislation shows areas of disagreement within state legislatures about how to regulate AI, as well as pushback from industry looking to ensure that a new technology is not stifled in its infancy. And while most states have adjourned their legislative sessions for the year, a handful of states are still in session, we’re likely to see some special legislative sessions called before the year is out, and New York’s governor still has significant AI bills pending on her desk.
Recent Developments
Major Policy Action
Colorado: Gov. Jared Polis (D) ordered lawmakers to return August 21 for a special session that would, among other things, address the artificial intelligence law that is set to go into effect in 2026. The order warns that it has become clear that the law, passed in 2024, “inadvertently imposes high costs on the state, local governments, and covered businesses” and requires solutions that reduce burdens “so that it meets the objectives of consumer protection and anti-discrimination while being simpler and less expensive to implement,” as well as pushing back implementation. One proposal would repeal the law in favor of a stripped-down law that prohibits discrimination by generative AI and requires disclosure to consumers upon interactions with AI.
California (1): The California Privacy Protection Agency board unanimously adopted a package of regulations related to automated decision-making technology (ADMT), cybersecurity audits, and risk assessments in late July. The regulations were stripped down this spring, but would still grant rights to consumers over the use of ADMT, including a right to be notified, a right to opt out, and a right to appeal for a “significant decision.”
California (2): On Tuesday, a federal judge struck down a political deepfake law in California challenged by Elon Musk and his X platform. The law would have blocked online platforms from hosting deceptive, AI-generated content within 120 days of an election, or 60 days afterward. The judge did not rule on First Amendment grounds, instead holding that the state law conflicted with federal law, namely Section 230 of the Communications Decency Act, which absolves platforms of liability for posted content.
Illinois: Gov. J.B. Pritzker (D) signed a bill into law (IL HB 1806) that prohibits offering therapy services, including AI-based services, without a licensed professional. It would also regulate the use of AI by therapists in transcribing sessions, requiring informed consent from the patient.
New York: Last Friday, Gov. Kathy Hochul (D) announced the first Empire AI supercomputer projects from the University at Albany. Empire AI is a $500 million consortium of leading public and private universities to advance responsible artificial intelligence.
Notable Proposals
Pennsylvania: House Democrats introduced a measure (PA HB 1779) that would require disclosures to consumers for using personalized algorithmic pricing. The proposal would also prohibit using protected class data to set prices.
Texas: During the special session, Rep. Dade Phelan (R) introduced a bill (TX HB 142) to prohibit political advertising that includes deepfakes without a disclosure. Phelan authorized a similar bill (TX HB 366) during the regular session that passed the House, but stalled out in the Senate.
