AI system disclosure to end users
ai-system-disclosureDomain: ai-transparencyType: policyDescription
AI disclosure obligations have a peculiar shape: most of them are not about the AI, they are about whether the user knows they are talking to one. EU AI Act Article 50 does not regulate the model; it regulates whether the natural person interacting with the system has been told it is a system. California SB 942 does not regulate what a generative model may produce; it regulates whether the output carries a label saying it was generated. FTC enforcement of deceptive AI advertising follows the same logic: the deception is the missing disclosure, not the underlying capability. The premise across the family of statutes is that user awareness, not model behavior, is the policy lever. A working AI-disclosure program has four operational pieces. The taxonomy comes first: which surfaces across the product actually involve AI in a way that triggers a disclosure obligation, and at what depth of involvement (an AI-drafted email looks different from an AI agent autonomously sending one). The disclosure surface comes second, and most teams underestimate how many of these are needed: a first-interaction modal handles new users, a help-center page handles auditors and journalists, a per-feature label in settings handles ongoing context, and a footer or sub-line handles AI-generated marketing copy. The accessibility layer comes third, because a disclosure that does not survive screen-reader rendering or a mobile reflow is not a disclosure under most of the relevant statutes. And the delivery log comes fourth: regulators have increasingly asked operators to show that the disclosure was actually displayed to the user who took the action, not just that the disclosure copy exists in the codebase. The thresholds are converging on the calendar. EU AI Act Article 50 obligations for AI systems interacting with natural persons take effect 2026-08-02. California SB 942 takes effect 2026-01-01 for generative AI providers above the million-monthly-user threshold. California AB 1836 (deceased-likeness deepfake disclosure) and the FTC's evolving guidance on AI-advertising claims are already in force. The substantive obligations differ across the four statutes; the disclosure obligation itself does not, which is why operators usually consolidate to a single disclosure surface and let regulator-specific copy fragments compose into it. The piece that consistently surprises operators is the interaction with marketing claims. A product page that promises an AI-powered feature creates an FTC-cognizable representation about that feature's behavior, and a misalignment between the marketing claim and the AI's actual capability is increasingly being pursued as deceptive advertising rather than as AI regulation per se. The disclosure-system owner therefore tends to be a cross-functional role spanning product, legal, and marketing, not a pure product role; the same disclosure framework that satisfies Article 50 also has to align with what the homepage and the App Store description say about the product.
Applicability
Applies when: AI role is ai-provider or ai-deployer.
Required by (12 regulations)
- EU AI Act
Article 50 imposes transparency obligations on providers and deployers of AI systems that interact with natural persons, generate or manipulate image / audio / video content, or perform emotion recognition or biometric categorization. Natural persons must be informed they are interacting with an AI system unless that fact is obvious from context.
Regulation (EU) 2024/1689 of the European Parliament and of the Council (Artificial Intelligence Act); Article 50 transparency obligations effective 2026-08-02
- California SB 942
Generative AI providers above the monthly-user threshold must offer a free AI-detection tool and apply latent and manifest disclosures to AI-generated content. The manifest disclosure is the user-facing label; the latent disclosure is the machine-readable provenance signal carried in the content metadata.
California SB 942 (California AI Transparency Act); effective 2026-01-01
- FTC AI Advertising
FTC guidance treats undisclosed AI involvement in consumer-facing interactions as a potentially deceptive practice under Section 5 of the FTC Act, particularly where the AI substitutes for a human role the consumer reasonably expected. Marketing claims about AI capability are evaluated under the same substantiation standard as any other product claim.
FTC enforcement guidance on AI claims and AI-mediated consumer interactions under Section 5 of the FTC Act, 15 U.S.C. § 45
- California AB 1836
AB 1836 prohibits the commercial distribution of digital replicas of deceased personalities without consent of the rights-holder, and requires that AI-generated likeness content carry a disclosure identifying it as such. The right-of-publicity overlay on generative AI makes the disclosure obligation operational rather than aspirational.
California AB 1836; amends Civil Code section 3344.1 (digital replicas of deceased personalities)
- Illinois AIVIA
Provides the pre-interview notice that AI may be used to analyze the applicant's video interview (820 ILCS 42/5).
Illinois Artificial Intelligence Video Interview Act
- California SB 243
Provides the clear and conspicuous AI-identity disclosure SB 243 requires when a reasonable person would otherwise be misled (§ 22602).
California SB 243 (2025), Companion Chatbots
- California SB 1001
Provides the clear and conspicuous bot disclosure SB 1001 requires for bots used to incentivize a commercial transaction or influence a vote (§ 17941).
California SB 1001 (2018), B.O.T. Act
- California AB 2013
California AB 2013 generative-AI training-data transparency pairs with user-facing disclosure that an AI system is in use.
California AB 2013 (2024), codified at Cal. Bus. & Prof. Code §§22757 et seq.
- Colorado AI Act
The Colorado AI Act requires consumers be informed when interacting with an AI system in a consequential-decision context.
C.R.S. §§ 6-1-1701 et seq. (SB 26-189, repealing and replacing SB 24-205 (2024))
- Connecticut PA 23-16
Connecticut's AI provisions require disclosure that a consumer is interacting with an AI system.
Connecticut Public Act 23-16 (2023 Session); An Act Concerning Artificial Intelligence, Automated Decision-Making and Personal Data Privacy. Originally introduced as Senate Bill 1103. Signed by Governor Ned Lamont 2023-06-07.
- NYC LL144
NYC Local Law 144 requires candidates be notified that an automated employment decision tool is used in hiring or promotion.
NYC Local Law 144 of 2021; codified at NYC Admin. Code § 20-870 et seq.; implementing rules at 6 RCNY § 5-300 et seq.
- Utah AI Policy Act
The Utah AI Policy Act requires disclosure of generative-AI use on consumer request and in regulated occupations.
Utah Code §§ 13-2-12 (AI consumer-protection disclosure), 13-72-101 et seq. (Artificial Intelligence Policy Act and AI Learning Laboratory)
Evidence formats
- Terms of Service or UI disclosure section naming the AI system in use
- first-interaction modal or banner copy disclosing AI involvement
- help-center page describing the AI features and their role
- in-product per-feature AI labels (settings, tooltips, footers)
- sample disclosure messages or screenshots from the live product
- marketing-claim review log aligning product-page copy with AI capability