California AI laws
What California has enacted on AI, what each law requires, and what the bills still moving would do. California is one of 44 states that enacted an AI law in the 2025 and 2026 sessions.
188 new California laws take effect in 2027, 22 on AI
- Bills in scope
- 86bills in scope
- Past committee
- 74past committee
- Most recent action
- 30 September 2026most recent action
AI laws enacted in California
34 AI laws enacted in the California Legislature in the 2025 and 2026 sessions, newest first. Each links to the legislature's own page.
What it requires: Employers must not use workplace surveillance tools to monitor or surveil employees in bathrooms located in the workplace unless directed by court order. Employers must allow employees to leave behind workplace surveillance tools when entering bathrooms, except when employees must remain available during meal or rest periods or when tools are required for identification, safety, or bathroom access and meet specified conditions.
Employment, Law enforcement and criminal justice
What it requires: An employer shall not use a workplace surveillance tool that uses artificial intelligence to recognize or make inferences about an individual's emotional state, or to collect neural data. Employers may use workplace surveillance tools for safety purposes or if they do not meet these specifications.
Employment
What it requires: Health facilities, clinics, physician's offices, and group practice offices must take reasonable steps to ensure licensed health care providers retain independent professional judgment when care is informed by clinical decision support systems. These entities must not use AI tools to direct unlicensed personnel or independently perform clinical functions requiring professional licensure.
Health care
- AB 2392Public postsecondary education: generative artificial intelligence systems: procurement standards: training.Takes effect 1 January 2027
What it requires: The office of the Chancellor of the California Community Colleges and California State University must jointly convene a working group to develop GenAI procurement standards and training by January 1, 2028, provide that training to students, faculty, and staff, review and update training annually, and submit reports within 60 days of executing systemwide GenAI contracts.
Education
What it requires: A large online platform must detect provenance data in distributed content, provide a user interface indicating whether content was generated or altered by a GenAI system, allow users to inspect system provenance data, and must not knowingly strip system provenance data or digital signatures that comply with established standards-setting body specifications.
Transparency and disclosure
What it requires: A covered provider of a generative AI system publicly accessible in California must include a latent disclosure in AI-generated or altered image, video, or audio content conveying the provider's name, system name and version, creation/alteration time and date, unique identifier, and whether the system created or altered the content. Covered providers must make available a free disclosure verification tool. Covered providers must not falsely represent that a GenAI system is designed to primarily function as assistive technology.
Transparency and disclosure
What it requires: Any person is prohibited from knowingly using another's name, voice, signature, photograph, or likeness, including a digital replica, in any manner on products, merchandise, goods, or for purposes of advertising or selling without prior consent. False impersonation of another using a digital replica with intent to impersonate is prohibited.
Likeness and digital replicas
What it requires: A person must not knowingly use artificial intelligence to falsely represent that a natural person appeared before, submitted information to, or otherwise engaged with a governmental agency. Governmental agencies may use disclosure verification tools to determine if artificial intelligence is present in engagement with transparency and governance laws.
Transparency and disclosure, Government use
What it requires: Manufacturers of autonomous vehicles operating without a human operator must maintain dedicated emergency response telephone lines answered within 30 seconds, equip vehicles with two-way voice communication devices, maintain communications capacity for concurrent remote assistance sessions, ensure remote drivers are US-located with valid California licenses, immediately dispatch local incident technicians upon crashes or emergency requests, and maintain quarterly records on response times and incidents.
Autonomous vehicles
What it requires: Developers and deployers of clinical decision support systems must make reasonable efforts to identify systems with known or reasonably foreseeable risk of biased impacts. Developers must mitigate such risks and provide deployers with statements of intended uses, risks, training data summaries, performance evaluations, and monitoring recommendations. Deployers must regularly monitor systems and take reasonable steps to mitigate biased impacts.
Transparency and disclosure, Health care
- SB 574Attorneys, arbitrators, judicial officers, and alternative resolution providers.Takes effect 1 January 2027
What it requires: Attorneys must not delegate the practice of law to generative artificial intelligence, must not enter confidential or nonpublic information into generative AI systems without restricted access, must verify accuracy of all AI-generated citations, must correct erroneous AI outputs, and must disclose AI use to courts. Arbitrators must not delegate decisionmaking to generative AI and must disclose any reliance on AI-generated information outside the record.
Transparency and disclosure
What it requires: Beginning July 1, 2027, employers must not use automated decision systems to violate labor laws, infer protected status, or predict adverse action for exercising legal rights. Employers cannot rely solely on ADS for disciplinary or termination decisions and must have humans corroborate ADS output using supporting information. Employers must provide affected employees written notice and a description of their data used in such decisions.
Employment
What it requires: Employers at covered establishments must include in mass layoff, relocation, or termination notices caused by AI or automated technology: the number and classification of affected positions, job functions to be automated, the specific type of AI or automation technology, and the statement "This notice is for a technology displacement" at the top. The Employment Development Department must publish summaries of these notices on its website and post quarterly statewide summaries.
Employment
What it requires: Large private businesses must not represent customer service chatbots as human and must disclose when chatbots are artificially generated if a reasonable person would be misled. They must provide a feature allowing customers to request connection to a customer service agent during regular business hours and make good faith efforts to connect within 15 minutes or schedule an appointment within one business day.
Chatbots and companion AI
What it requires: A person, committee, or other entity shall not knowingly distribute materially deceptive content portraying candidates, elected officials, elections officials, or voting equipment as doing or saying something they did not do or say, with malice, during specified periods before and after elections. Candidates may distribute such content if they include required disclosures about manipulation.
Elections and political ads
What it requires: A person must not create and cause to be published in an advertising medium an advertisement that prominently includes a synthetic performer without a clear and conspicuous disclosure that the advertisement includes a synthetic performer. An advertising medium must not transmit, distribute, display, air, or otherwise make available an advertisement containing a synthetic performer after receiving a court order finding a violation and being served with that order.
Transparency and disclosure, Likeness and digital replicas
What it requires: Operators of internet websites, online services, online applications, or mobile applications designed or marketed for K-12, preschool, prekindergarten, or higher education purposes must not use covered information including persistent unique identifiers to train generative AI systems, engage in targeted advertising based on such information, sell pupil information, or collect reproductive health, immigration status, or sexual orientation data.
Education
What it requires: The State Department of Education must develop a plan by January 1, 2028 to expand digital wellness instruction in California public schools. The plan must include age-appropriate instructional resources for elementary through high schools, strategies for curriculum currency, guidance for under-resourced schools, family engagement components, assessment methods, and educator professional development strategies.
Education
What it requires: The Instructional Quality Commission shall consider incorporating cybersecurity skills content into the computer science content standards when those standards are next revised after January 1, 2027. The commission shall also consider incorporating media literacy and AI literacy into specified curriculum frameworks and instructional materials adoption criteria by stated dates.
Education
What it requires: Operators of companion chatbots must perform comprehensive risk assessments for child users, implement crisis response protocols, usage limits, parental controls, and safety features by July 1, 2027. Operators must submit to independent child safety audits every two years and prohibit cross-context behavioral advertising and sale of child personal information.
Chatbots and companion AI
What it requires: No person shall manufacture, sell, exchange, possess with intent to sell or exchange, or expose or offer for sale or exchange to any retailer any toy that includes a companion chatbot, until January 1, 2031.
Chatbots and companion AI
What it requires: Beginning January 1, 2029, persons must register with the Government Operations Agency before offering, selling, or conducting a covered AI audit. Registered AI auditors must provide specified information, conduct audits according to industry standards, deliver reports with signed statements, maintain independence and objectivity, and retain audit documentation for 10 years.
Comprehensive AI, Government use
What it requires: On or before January 1, 2028, the Government Operations Agency must develop application requirements, suspension and termination procedures, and designation criteria for independent verification organizations. The agency must convene working groups, consult stakeholders, publish requirements on its website, and regularly review criteria.
Government use
What it requires: The State Bar of California must disclose on its internet website 60 days before each examination the use of artificial intelligence-generated content in developing or administering State Bar examinations. The State Bar must disclose artificial intelligence-generated content in study materials on the cover page. These disclosures apply to content developed by or at the explicit direction of the State Bar, regardless of whether the content is revised or reviewed by a natural person.
Task forces and studies
What it requires: A defendant who developed, modified, or used artificial intelligence may not assert as a defense that the artificial intelligence autonomously caused harm to the plaintiff. Defendants may still present other affirmative defenses and evidence regarding causation, foreseeability, and comparative fault.
Transparency and disclosure
What it requires: A person must not create and intentionally disclose, or intentionally disclose, digitized sexually explicit material of a depicted individual without consent or when the depicted individual was a minor. A person providing services enabling a deepfake pornography service's operation must cease doing so within 30 days of receiving evidence of violation.
Intimate images and CSAM
What it requires: Beginning January 1, 2027, large online platforms must detect provenance data compliant with established standards, provide user interfaces disclosing system provenance data, and allow users to inspect such data. Beginning January 1, 2028, capture device manufacturers must provide users an option to include latent disclosures in captured content and embed such disclosures by default.
Transparency and disclosure
What it requires: Operators of companion chatbot platforms must issue clear notification that chatbots are artificially generated if a reasonable person would be misled into believing they are human. Operators must maintain protocols preventing suicidal ideation, suicide, or self-harm content and publish protocol details online. For minor users, operators must disclose artificial intelligence use and provide reminder notifications every three hours.
Chatbots and companion AI
- AB 489Health care professions: deceptive terms or letters: artificial intelligence.Enacted 11 October 2025
What it requires: Entities that develop or deploy artificial intelligence or generative artificial intelligence systems must not use terms, letters, or phrases in advertising or functionality that indicate or imply the advice, care, reports, or assessments are provided by a natural person with an appropriate health care license or certificate.
Transparency and disclosure, Health care
What it requires: Each law enforcement agency must maintain a policy requiring official reports generated using artificial intelligence to identify every AI program used and prominently state AI was used, include the officer's signature verifying accuracy, retain the first draft as long as the official report is retained, and maintain an audit trail identifying who used the AI and any video or audio footage used.
Law enforcement and criminal justice
What it requires: A person must not use or distribute a common pricing algorithm as part of a contract, combination, trust, or conspiracy to restrain trade or commerce. A person must not use or distribute a common pricing algorithm if coercing another person to set or adopt a recommended price or commercial term for the same or similar products or services in California.
Algorithmic pricing
What it requires: Manufacturers of autonomous vehicles must submit applications to the Department of Motor Vehicles and obtain approval before operating on public roads. Commencing January 1, 2026, autonomous vehicles may be equipped with ADS marker lamps. Commencing January 1, 2030, autonomous vehicles with model year 2031 or later and gross vehicle weight rating under 8,501 pounds must be zero-emission vehicles.
Autonomous vehicles
What it requires: The California Cybersecurity Integration Center must develop a California AI Cybersecurity Collaboration Playbook by January 1, 2027, in consultation with the Office of Information Security and the Government Operations Agency. The playbook must include mandatory information-sharing mechanisms for state contractors and vendors providing AI services regarding threats and vulnerabilities. Information shared under the playbook is confidential and shall not be transmitted except to approved state employees and contractors.
Government use, Critical infrastructure
What it requires: Large frontier developers must write, implement, and publish frontier AI frameworks describing their approach to catastrophic risk management. They must publish transparency reports before deploying new or substantially modified frontier models. Large frontier developers must transmit quarterly summaries of catastrophic risk assessments from internal model use to the Office of Emergency Services. Frontier developers must report critical safety incidents within 15 days, or 24 hours if imminent risk exists.
Frontier and general-purpose AI
AI in California, by topic
Transparency and disclosure
15 bills, 9 enacted. Enacted in 9 states.
Employment
10 bills, 4 enacted. Enacted in 4 states.
Government use
10 bills, 4 enacted. Enacted in 9 states.
Chatbots and companion AI
8 bills, 4 enacted. Enacted in 15 states.
Education
8 bills, 4 enacted. Enacted in 12 states.
Health care
6 bills, 3 enacted. Enacted in 10 states.
Likeness and digital replicas
6 bills, 2 enacted. Enacted in 13 states.
Autonomous vehicles
6 bills, 2 enacted. Enacted in 10 states.
Task forces and studies
6 bills, 1 enacted. Enacted in 12 states.
Algorithmic pricing
5 bills, 1 enacted. Enacted in 4 states.
Law enforcement and criminal justice
4 bills, 2 enacted. Enacted in 8 states.
Critical infrastructure
4 bills, 1 enacted. Enacted in 4 states.
Comprehensive AI
2 bills, 1 enacted. Enacted in 5 states.
Frontier and general-purpose AI
2 bills, 1 enacted. Enacted in 2 states.
Intimate images and CSAM
2 bills, 1 enacted. Enacted in 19 states.
Automated decision-making
2 bills. Enacted in 2 states.
Insurance
2 bills. Enacted in 10 states.
Elections and political ads
1 bill, 1 enacted. Enacted in 10 states.
AI bills in California
The bills not yet law, furthest along first, with what each would do.
34Enacted
9Cleared both chambers
31Past committee
12Introduced or inactive
86 bills in scope
Cleared both chambers
9
What it would do: Prohibits employers from retaliating against healthcare workers for using professional judgment to override clinical decision support system outputs, and prevents defendants from claiming such worker override is a superseding cause eliminating liability for AI system-caused harm.
What it would do: Requires public employers identified in Government Code Section 3555.5 to provide recognized employee organizations at least 45 days' written notice before developing, purchasing, or requiring use of generative artificial intelligence to perform services within the scope of work of represented job classifications.
What it would do: Regulates artificial intelligence use by mental health providers, permitting AI only for administrative or supplementary support, and prohibiting AI from making therapeutic decisions, directly communicating with patients, or offering psychotherapy via chatbots without patient consent and licensed professional review.
What it would do: Requires AI providers to warn consumers by December 1, 2026 that unlawful use of digital replica technology may cause civil or criminal liability, with violations subject to up to $10,000 daily penalties. Clarifies that digital replicas constitute impersonation and voice/likeness use under existing law, and directs the Judicial Council to develop court rules for AI-generated evidence by January 1, 2027.
What it would do: Requires employers to notify workers about automated decision systems used for employment decisions, prohibits certain ADS uses, mandates human review for discipline/termination decisions, and allows workers to request their data. Establishes Labor Commissioner enforcement with $500 civil penalties.
What it would do: Makes social media platforms with over $100 million annual revenue liable for civil penalties up to $1,000,000 if they violate California civil rights laws through algorithms that relay content or by aiding violations, effective January 1, 2027.
What it would do: Prohibits operators from making companion chatbots available to children under 18 unless the chatbot cannot foreseeably encourage self-harm, suicidal ideation, violence, drug/alcohol use, disordered eating, or engage in sexually explicit interactions. Allows the Attorney General and harmed children to pursue civil actions for violations.
What it would do: Shortens prior authorization decision timelines for health care service plans and disability insurers to 3 business days for electronic submissions or 5 business days for non-electronic submissions, and to 24 hours for electronic or 48 hours for non-electronic submissions in urgent cases. Excludes Medi-Cal managed care plans from these timeline changes.
What it would do: Requires health care service plans and health insurers that impose prior authorization to publicly report prior authorization data on their websites and annually report claims and prior authorization data to regulators, with regulators posting disaggregated information online.
Past committee
16 of 31 shown
What it would do: Requires developers and deployers of automated decision systems used for consequential decisions to provide notice, documentation, and appeal opportunities to affected individuals. Authorizes the Attorney General and other officials to enforce compliance with civil penalties up to $25,000 per violation.
What it would do: Requires companion chatbot operators to perform annual child safety risk assessments, implement protective measures including parental controls and usage limits, submit to independent audits, and prohibits certain harmful practices. Allows civil enforcement by prosecutors and private right of action for harmed children.
What it would do: Requires manufacturers of advanced driver assistance systems to allow drivers to disable or disengage the system as a condition of operating the vehicle where manual operation was previously available. Authorizes civil penalties up to $25,000 per violation and requires DMV to consider adding questions about advanced driver assistance systems to driver's license examinations.
What it would do: Amends companion chatbot regulations to require operators to prevent production of any sexually explicit material when they have constructive knowledge a user is a minor, and creates an exception allowing public officials to make decisions affecting nonprofit organization members without disclosing financial interests.
What it would do: Expands language access requirements for state agencies by revising the definition of 'substantial number' to include people eligible to be served, requiring comprehensive language needs assessments using census data, and establishing a Language Access Director within the California Health and Human Services Agency to oversee language access plans.
What it would do: Creates a tax credit for qualified taxpayers producing commercials in California for taxable years 2027-2031, allowing 20-30% of qualified production costs exceeding $500,000, with $15,000,000 annual allocation limit and restrictions on generative AI and autonomous vehicle use.
What it would do: Establishes the California Artificial Intelligence Worker Impact Data Assessment Project and a 14-member advisory panel within the Employment Development Department to assess data sources on artificial intelligence's impact on the labor force and submit policy recommendations to the Legislature by January 1, 2028.
What it would do: Creates a crime for adults who recruit, coerce, or persuade minors to commit illegal acts against other minors or solicit sexual conduct or intimate images from them, punishable by up to one year in county jail or $1,000 fine. Also removes the extortion law exemption for minors under 18 who threaten or coerce other minors for sexual conduct or intimate images.
What it would do: Defines agentic artificial intelligence and requires the Department of Technology to inventory high-risk automated decision systems and agentic AI used by state agencies, and requires the Office of Emergency Services to include agentic AI in risk analyses of threats to critical infrastructure.
What it would do: Requires advertisements depicting digital replicas or synthetic performers as health care providers promoting health-related consumer products to include clear, conspicuous disclosure that the provider was AI-generated or no human provider is depicted. Violations constitute unfair competition and false advertising violations.
What it would do: Requires state agencies operating critical infrastructure to establish human oversight mechanisms for AI systems by July 1, 2026, conduct annual safety assessments, and report AI adverse events within specified timeframes, with civil penalties up to $500 per seven days for non-compliance.
What it would do: Encourages the Presidents and Chancellors of UC, CSU, and California Community Colleges to create a workgroup to review artificial intelligence usage in higher education and submit a public report with strategies and best practices to the Legislature within one year.
What it would do: Requires operators of companion chatbots to adopt policies for detecting credible crisis expressions, implement monitoring systems, provide warnings and crisis interruption pauses of 20 minutes upon detection, and annually report data to the Office of Suicide Prevention beginning January 1, 2028.
What it would do: Requires state agencies using generative AI to directly communicate with the public to include a disclaimer indicating the communication was AI-generated and provide information on contacting a human employee, with specific placement requirements based on communication type.
What it would do: Requires the Public Utilities Commission to adopt AI safety standards for electrical and gas corporations by January 1, 2028, including disclosure, human review requirements, and workforce protections. Requires community choice aggregators and local publicly owned electric utilities to adopt consistent AI policies.
What it would do: Requires employers to provide written notice to workers about workplace AI tools used for employment decisions or surveillance, including specific information about the tool's purpose, data collected, and job impacts. Establishes enforcement by the Labor Commissioner or public prosecutors, with penalties up to $500 per violation.
Showing the 25 furthest along of 52. Subscribers see all of them, filtered and searchable. Bill titles link to the California Legislature's own page.
What counts as AI legislation
In scope is any bill that puts a duty on someone who builds, sells or deploys an automated system: disclosure when a model generates content, testing and notice before an algorithm screens a job applicant or sets a price, limits on government facial recognition, and consent for a synthetic likeness. The tracker follows the mechanism rather than the industry, so an insurance bill and a hiring bill arrive together when both create the same obligation.
Questions
Does California have an AI law?
Yes. California enacted 34 AI laws in the 2025 and 2026 sessions, among them AB 1331, AB 1883 and AB 1979. Each is listed on this page with what it requires.
When do California's AI laws take effect?
AB 1331 on 1 January 2027, AB 1883 on 1 January 2027, AB 1979 on 1 January 2027, AB 2392 on 1 January 2027 and AB 2713 on 1 January 2027. The others have no single date in their text.
Which AI bills are moving in California?
40 bills not yet law have passed a committee or a chamber, the furthest along AB 2575 (vetoed), AB 2656 (vetoed) and SB 903 (vetoed).
What does California legislate on most in AI?
Transparency and disclosure (15 bills), Employment (10 bills) and Government use (10 bills).
Are these summaries reviewed by a lawyer?
No. Which bills count as AI bills, and the line on what each law requires or each bill would do, are drafted by software from the bill's own text and not reviewed by an attorney. Every bill links to the legislature's own record: read the text before relying on it. The list covers bills from the 2025 and 2026 sessions.
Other practice areas in California
- Alcohol in California33
- Cannabis in California19
- Children's Online Safety in California22
- Cybersecurity in California9
- Data center and energy in California15
- Data Privacy in California61
- Employment in California208
- Entertainment in California23
- Healthcare in California276
- Insurance in California153
- Intellectual Property in California11
- Packaging and PFAS in California35
- Payments and fintech in California15
- Rental housing in California42
- Tobacco and nicotine in California8
- Transportation in California27
- All jurisdictions
AI legislation in other jurisdictions
52 more currently hold bills in this tracker
- Congress377
- New York178
- New Jersey163
- Illinois98
- Texas90
- Virginia65
- Pennsylvania61
- Maryland58
- Hawaii54
- Massachusetts51
- Missouri40
- Iowa39
- Montana39
- Connecticut36
- Louisiana34
- Tennessee34
- Florida33
- Puerto Rico32
- Mississippi31
- Georgia30
- Oklahoma30
- Vermont29
- Minnesota28
- North Carolina28
- Utah27
- Washington27
- Colorado23
- Rhode Island23
- Arizona21
- Nevada20
- Ohio20
- Kentucky19
- Michigan17
- New Mexico17
- Wisconsin17
- West Virginia17
- South Carolina16
- Arkansas14
- Maine14
- Nebraska11
- New Hampshire11
- Oregon11
- Alabama10
- Idaho10
- Indiana8
- Kansas8
- District of Columbia7
- North Dakota7
- South Dakota7
- Delaware6
- Alaska5
- Wyoming4
Know when California's AI law changes
An alert when a bill moves, and one report a month on what changed in every state and Congress.