AI disclosure laws

Where AI-content disclosure is the law

A cited, plain-language map of the world's AI-content-disclosure and AI-interaction laws: what each one requires, who it binds, and when it takes effect. AIDisclose declares, it never detects: we make no claim about whether content is AI. This directory states what is legally required per region and where adoption is voluntary best practice. It is a general reference and does not provide legal advice. Always confirm with your own counsel.

All entries reviewed against primary sources as of 5 July 2026.

Binding disclosure laws

In force or scheduled. AIDisclose maps these to concrete disclosure obligations.

European Union

Binding law

EU AI Act (Regulation (EU) 2024/1689), Article 50 Effective 2 August 2026

Required: Providers must disclose AI interaction (50(1)) and mark synthetic audio/image/video/text as machine-readable (50(2)). Deployers must notify people exposed to emotion-recognition or biometric-categorisation systems (50(3)) and label deepfakes and AI-generated text published to inform the public on matters of public interest (50(4)).

Binding from 2 August 2026 (firm, per Regulation (EU) 2024/1689). A short transitional period for the 50(2) machine-readable marking of pre-existing generative-AI systems was set by the AI/Digital Omnibus amending regulation (final text adopted 29 June 2026); it is reported as 2 December 2026 by the European Parliament's Legislative Train, but is not yet published in the Official Journal and the original Commission proposal said 2 February 2027 — treat the exact transitional date as provisional until OJ publication. A voluntary Code of Practice on Transparency of AI-Generated Content was published by the EU AI Office on 10 June 2026.

Applies to: Providers and deployers of AI systems placing outputs on the EU market.

United States — California

Binding law

California AI Transparency Act (SB 942, amended by AB 853) Effective 2 August 2026

Required: Covered providers must embed latent (machine-readable) provenance disclosure in AI-generated image/video/audio, offer users a clear user-visible label option, and make available a free public AI-detection tool.

Operative for covered providers from 2 August 2026, deliberately aligned with the EU AI Act. Large online platforms and GenAI hosting platforms phase in 1 January 2027; capture-device manufacturers 1 January 2028. California also has SB 243 (companion chatbots, 1 Jan 2026), AB 2013 (training-data transparency, 1 Jan 2026), and SB 1001 (bot disclosure, in force since 2019).

Applies to: Covered providers of publicly accessible generative AI systems with more than 1,000,000 monthly users in California.

United States — Texas

Binding law

Texas Responsible AI Governance Act (TRAIGA, HB 149) Effective 1 January 2026

Required: Government agencies must clearly and conspicuously disclose to consumers that they are interacting with an AI system, before or at the point of interaction. Healthcare providers must disclose use of AI in a patient's treatment.

In force since 1 January 2026. Enforcement is exclusive to the Texas Attorney General with a 60-day cure period and no private right of action.

Applies to: Government agencies and healthcare providers/entities. Private businesses have no general AI-interaction disclosure duty.

United States — Utah

Binding law

Utah AI Policy Act (SB 149, amended by SB 226) Effective 1 May 2024

Required: Must disclose that a consumer is interacting with generative AI (not a human) on a clear and unambiguous request, and proactively at the outset of high-risk interactions that both collect sensitive personal information and provide personalized advice for significant decisions.

In force since 1 May 2024, narrowed by SB 226 in 2025; sunset extended to 1 July 2027. A safe harbour applies for entities that clearly disclose GenAI use throughout the interaction.

Applies to: Any person using generative AI in a consumer interaction.

United States — New York

Binding law

New York General Business Law Article 47 (S3008) Effective 5 November 2025

Required: Operators must clearly disclose that the user is engaging with AI and not a human at the start of interaction and at least every three hours of continuing interaction, plus self-harm detection and referral duties.

In force since 5 November 2025. Attorney-General enforced, up to $15,000/day.

Applies to: Operators of AI companion models.

China

Binding law

CAC Measures for Labeling AI-Generated Synthetic Content + GB 45438-2025 Effective 1 September 2025

Required: Content must carry explicit labels (user-perceptible text/audio/visual markers) and implicit labels (machine-readable metadata/watermarks embedded in the file). Users must not remove or forge labels.

In force since 1 September 2025, building on the 2023 Deep Synthesis Provisions.

Applies to: Generative-AI service providers, content-dissemination platforms, app-distribution platforms, and end users.

Italy

Binding law

Italian Law No. 132/2025 Effective 10 October 2025

Required: Requires transparency and labelling of AI-generated or altered content and disclosure when AI assists decisions in sensitive areas; creates a criminal offence for the illicit dissemination of harmful AI-altered content.

First national AI law inside the EU, in force since 10 October 2025. Complements, and does not override, the directly applicable EU AI Act.

Applies to: Providers and deployers operating in Italy (complements the EU AI Act).

South Korea

Binding law

Framework Act on AI (AI Basic Act) + Enforcement Decree Effective 22 January 2026

Required: Operators must notify users in advance that AI is in use and, where outputs are hard to distinguish from real content, label outputs as AI-generated (human-perceptible or machine-readable; a one-time human-readable notice is required either way).

In force since 22 January 2026. Administrative fines and fact-finding investigations are deferred for roughly one year in 2026 except in cases of serious harm.

Applies to: Operators providing products or services using generative or high-impact AI.

India

Binding law

IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 Effective 20 February 2026

Required: Mandatory prominent labelling of synthetically generated information at upload, unalterable embedded metadata/identifiers, a visible on-screen label and audible warning for audio, plus fast takedown windows.

In force since 20 February 2026 (notified by MeitY on 10 February 2026).

Applies to: Intermediaries and platforms.

Proposed laws

Bills that would create disclosure duties but are not yet in force.

United States — Colorado

Proposed

Colorado SB 26-189 (successor to the repealed SB 24-205) Effective 1 January 2027

The original Colorado AI Act (SB 24-205) was repealed before taking effect. Its successor SB 26-189 focuses on automated-decision-making notices and takes effect 1 January 2027 — it is not a general AI-content labelling regime.

Applies to: Deployers and developers of automated decision-making technology.

Spain

Proposed

Draft Organic Law on the Good Use and Governance of AI

A bill in parliamentary processing (submitted May 2026), expected to enter into force around end of 2027, with fines up to €35 million or 7% of global turnover. Until then, the directly applicable EU AI Act is what is enforceable in Spain.

Applies to: Providers and deployers operating in Spain (the directly applicable EU AI Act governs meanwhile).

Brazil

Proposed

PL 2338/2023

A risk-based bill modelled on the EU AI Act, approved by the Senate on 10 December 2024 and pending in the Chamber of Deputies. Generative-AI provisions would take effect 180 days after enactment.

Applies to: Providers and deployers of AI systems (once enacted).

No binding law, voluntary adoption

No statutory AI-disclosure duty today. Publishers adopt the standard as best practice and to get ahead of regulation.

United States (federal)

No binding law

No federal AI-disclosure statute in force

There is no enacted federal AI-content-disclosure law. AI disclosure in the US is governed by state laws (e.g. California SB 942, Texas TRAIGA). Adopt the AIDisclose standard as best practice and to prepare for state obligations.

Canada

No binding law

AIDA (Bill C-27) lapsed at prorogation, 6 January 2025

There is no binding federal AI-transparency law after Bill C-27 (and AIDA) died on the Order Paper. Only a non-binding Voluntary Code of Conduct on generative AI exists. Adopt as voluntary best practice.

United Kingdom

No binding law

No horizontal AI-labelling law

The UK has no statutory AI-content-labelling or transparency-disclosure duty; it relies on existing sectoral regulators and adjacent measures (e.g. the Online Safety Act, and a non-consensual intimate deepfake offence). Adopt as voluntary best practice.

Global / other

No binding law

No jurisdiction-specific mandate asserted

Voluntary adoption of the AIDisclose transparency standard, independent of any single legal regime — a global best-practice baseline.

Checked 5 July 2026

Sources are linked per entry and re-checked periodically; see the date on each card. Found something out of date? The registry is open: read the spec or open an issue.