Summary
- Covered providers must, where technically feasible, embed a latent (hidden) disclosure in image, video or audio content their GenAI system creates or alters (other than minor edits such as cropping or brightness). It must carry the provider name, system name and version, time and date, a unique ID, and whether the content was created or altered, and be permanent or extremely hard to remove (BPC 22757.3(a)).
- Covered providers must offer a free, public disclosure verification tool (formerly called an "AI detection tool") that accepts uploads or URLs, has an API, shows system provenance data, and does not reveal personal data without consent. A compliant third-party tool can be used instead (22757.2).
- Licensing: providers must tell licensees about their duties; if a licensee strips compliance, the provider must revoke the licence or notify it within 72 hours; the licensee must fix or stop within 96 hours, or the provider reports it to the AG (22757.3(b)).
- From 1 Jan 2027: large online platforms (over 2 million unique monthly users) must detect standard provenance data, show users whether content is AI-generated or camera-captured, and not strip it; GenAI hosting platforms must not knowingly offer models that skip the disclosures (22757.3.1, 22757.3.2).
- From 1 Jan 2028: makers of cameras, phones and voice recorders first sold in California from that date must embed latent disclosures by default, with a user option (22757.3.3).
Who it applies to
- Covered provider: anyone who creates, codes or otherwise produces a GenAI system publicly accessible within California (22757.1(e), as amended by SB 1000).
- Extraterritorial: yes. The test is public accessibility in California, not where the company sits. An Indian image, video or voice generator open to California users is covered.
- Large online platforms, GenAI hosting platforms (sites offering model weights or source code for download) and capture device manufacturers have their own later duties.
- Exclusions: products offering only non-user-generated video games; until 1 Jan 2029, GenAI systems designed mainly as assistive technology for people with disabilities (22757.5).
Key dates
- SB 942 approved by Governor (Chapter 291); original operative date 1 Jan 2026
- AB 853 approved (Chapter 674): operative date moved to 2 Aug 2026; platform, hosting and device duties added
- Act operative for covered providers (BPC 22757.6)
- SB 1000 approved as urgency statute (Chapter 861), effective immediately: user threshold and manifest disclosure duty removed
- Large online platform and GenAI hosting platform duties startUpcoming
- Capture device duties start for devices first produced for sale from this dateUpcoming
- Assistive technology exemption ends; latent disclosure must flag assistive technology systemsUpcoming
Full date record
| Date | Event |
|---|---|
| 2024-09-19 | SB 942 approved (Chapter 291) |
| 2025-10-13 | AB 853 approved (Chapter 674), effective 1 Jan 2026; operative date moved to 2 Aug 2026 |
| 2026-08-02 | Covered provider duties operative |
| 2026-09-30 | SB 1000 approved and effective immediately (Chapter 861) |
| 2027-01-01 | Large online platform and GenAI hosting platform duties |
| 2028-01-01 | Capture device duties for newly produced devices |
| 2029-01-01 | Assistive technology exemption and USD 50,000 false-claim penalty section end |
Penalties and enforcement
- USD 5,000 per violation, collected by the AG, a city attorney or a county counsel; each day a covered provider, large online platform or device maker is in breach is a separate violation; winning plaintiffs get attorney fees (22757.4).
- USD 50,000 per violation, per day, for falsely claiming a GenAI system is assistive technology; this section repeals on 1 Jan 2029 (22757.4.1).
- Licensees who fail to fix or stop: injunction plus fees (22757.4).
- Lawsuits filed before SB 1000 cannot continue for conduct that is lawful under the amended text (22757.4(c)).
- No private right of action for consumers. No court challenge to the Act was found as of 2026-10-19 (limited search).
What it means for an Indian company
- Gen-AI image, video, avatar, voice and music apps with California users: you are likely in scope now, whatever your user count. Embed latent provenance (C2PA-style manifests and watermarks) on every output and test that it survives normal edits (SC-AI-023 Label AI-generated and synthetic content).
- Ship a verification tool or point to a compliant third-party one with upload, URL and API options, and keep it from leaking personal data (SC-AI-012 Data minimisation and retention for AI).
- White-label and API resellers: your licence terms need a compliance clause, a 72-hour revoke-or-notify process and an AG reporting path (SC-AI-007 Third-party and vendor AI due diligence, SC-AI-031 AI incident management and regulator reporting).
- Text-only chatbots are outside the disclosure duties, but still check other rules on AI interaction disclosure (SC-AI-022 Disclose AI interaction to users).
- Social, messaging and search platforms above 2 million monthly users need provenance detection and display by 1 Jan 2027 (SC-AI-036 Deepfake and impersonation safeguards).
- Indian phone and camera makers selling in California should plan default latent capture metadata for 2028 models.
- Penalties stack daily, so add the Act to your applicability register (SC-AI-009 Regulatory applicability register and horizon scanning).
Obligations
Full obligation breakdown coming
We break the most requested policies down into numbered obligations mapped to controls. Leave your details and we’ll tell you when California AI Transparency Act (SB 942) is done.
Glossary
- Covered provider: a maker of a GenAI system publicly accessible in California.
- Latent disclosure: hidden provenance data embedded in content, readable by tools.
- Manifest disclosure: a visible label; no longer required after SB 1000.
- Provenance data: data embedded in or attached to content showing its origin and edit history, in a format meeting widely adopted standards.
- System provenance data: provenance data that cannot reasonably be tied to a particular user.
- Disclosure verification tool: the free checker providers must offer.
- Large online platform: a social, file-sharing, mass messaging or search service with over 2 million unique monthly users.
- Capture device: a camera, phone with camera or microphone, or voice recorder.
- Urgency statute: a California law passed by two-thirds vote that takes effect immediately.
Key takeaways
- Correct operative date is 2 Aug 2026, not 1 Jan 2026 (AB 853 delayed it).
- SB 1000 (30 Sep 2026) removed the 1 million user threshold and the manifest disclosure option, and renamed the detection tool.
- Duties today: latent provenance on AI images, video and audio, a free verification tool, and licensee controls.
- Platforms join on 1 Jan 2027, capture devices on 1 Jan 2028.
- USD 5,000 per violation, counted per day; reaches any provider whose system is publicly accessible in California.
Official sources
- BPC Chapter 25, California AI Transparency Act, current code text, accessed 2026-10-19
- SB 942 (2024) chaptered text, accessed 2026-10-19
- AB 853 (2025) chaptered text, accessed 2026-10-19
- SB 1000 (2026) chaptered text, accessed 2026-10-19
This is general information, not legal advice. Last verified 19 Oct 2026 against the official source. Laws change; confirm the current position with qualified counsel before acting.