AI Disclosure Rules: When the law requires you to label AI
More and more, the law does not ask whether you used AI. It asks whether you said so. We track the rules that require a visible label on AI-altered media, AI-generated content, and automated decisions.
A topic page in AI Regulation News, our source-backed tracker of the rules that change how professionals use AI at work.
A distinct branch of AI regulation has nothing to do with whether AI is allowed and everything to do with whether you disclosed it. Across sectors, lawmakers and regulators have settled on the same lever: not a ban, but a labeling duty. If a listing photo was altered by AI, the rule says tell the buyer. If an applicant is screened by an automated system, the rule says tell the applicant. If a chatbot or generated image is presented to the public, the rule says make the artificial origin clear. The common thread is transparency as the price of using the tool, and it is one of the easiest duties to satisfy and one of the easiest to forget.
The pattern matters because disclosure rules are unusually enforceable. A vague reasonableness standard invites argument. A labeling requirement does not. Either the conspicuous notice is on the listing, in the applicant communication, or beside the generated image, or it is not. That binary quality is exactly why these rules carry teeth, including misdemeanor exposure for willful violations in some real estate marketing statutes and administrative discipline in others. The regulator does not have to prove intent to deceive. The absence of the label can be the violation.
What is at stake is sharpest where AI touches a consumer-facing representation. Real estate marketing is the clearest case: an altered or AI-staged image that changes what a property looks like can mislead a buyer, so the rule pairs a conspicuous disclosure with a link to or archive of the original. Employment is the next: applicant notice that AI is part of the decision. And the European model reaches widest, requiring that synthetic media and AI systems interacting with people be marked as artificial, a standard that follows any firm with clients or content in that market regardless of where the firm sits.
The workflow consequence is small to build and expensive to skip. The defensible practice is a standing disclosure step in any process that produces public-facing or decision-affecting output: a fixed label on altered media, an archived original you can produce, an applicant notice that matches the strictest jurisdiction in your funnel, and a clear marker on anything generated that a person will read or rely on. Wire it into the template once and it runs itself. The briefings below track the specific disclosure rules now in force and the ones arriving. This is workflow interpretation across professions, not legal advice; it sits beside our court-rules and hiring-laws pages, which cover the duties that are not about labeling.
The briefings we track in this lane
California AB 723 and the AI Listing Photo Disclosure Rule
Altered or AI-staged listing images now need a conspicuous disclosure plus the original, and willful violations carry misdemeanor exposure for agents.
Read the briefing StatuteEU AI Act Article 50 and the Transparency Duty Reaching Your Firm
Article 50 requires that synthetic media and AI that interacts with people be marked as artificial. Why it follows firms far outside the European Union.
Read the briefing RuleIllinois HB 3773: The Four AI Disclosures Every Applicant Needs
The IDHR draft rules name what an AI-use notice must say. A paste-ready four-part applicant disclosure template plus a deployment checklist.
Read the briefingBrowse the full tracker for every rule, ruling, and agency action we follow across professions.