If your content team has touched an AI tool in the past year, and most have, you are already operating under disclosure rules, whether or not anyone has told you. The EU AI Act's transparency obligations took effect August 2, 2026. The FTC has been enforcing its existing deception authority against AI-generated content for longer than that. China's labeling regime has applied since September 2025. None of this is on the horizon. It is the current environment, and it applies across three different legal frameworks that do not map onto each other.
Most of what marketing teams hear about these rules arrives secondhand, through headlines that flatten a fairly narrow legal test into a blanket "label everything" mandate, or through silence, because no one on the team has been asked to own the question. Both are expensive. Overcorrecting dilutes content that never needed a label. Under-correcting leaves you exposed to fines, FTC enforcement, or a platform takedown you did not see coming.
We wrote a full guide, The CMO's Guide to AI Content Disclosure, to give marketing leaders a straight answer to what is required, where the real judgment calls sit, and what to do about it in the next 90 days. A few things it covers that surprised us in the researching:
- The EU text rule is narrower than reported. Labeling AI-generated text is required only when three specific conditions are all met, and content that has gone through genuine human review is exempt entirely, no matter how much AI was involved in the draft. The guide walks through exactly what "genuine review" needs to look like to hold up.
- Deepfake disclosure is the wider net. Separate from the text rule, a broader duty covers AI-generated images, audio, and video that could pass as real, with fines reaching EUR 15 million or 3% of worldwide turnover.
- The US has no single AI disclosure law, and that is not the same as no exposure. The FTC is actively enforcing its existing deception authority and its rule against fabricated reviews, with civil penalties running to $53,088 per violation.
- China's labeling regime does not interoperate with the EU's. If your content reaches Chinese platforms, you likely hold a distinct publisher obligation regardless of what your tool vendor does.
- AI text is becoming detectable, even where it is not yet. Major providers are now watermarking model output. Public detection tools lag behind those marks today, but content published now can be identified later, against an archive that does not go away.
The guide also includes a scenario-by-scenario decision table (blog posts, campaign imagery, synthetic voice, chatbots, and more, mapped against EU, US, and voluntary standards) and a ten-action checklist for building a disclosure program in the next quarter, however far along your team already is.
The short version: the regulatory floor is more workable than most commentary suggests, and a well-run editorial process already gets you most of the way there. The harder question, the one the guide spends its last section on, is whether the floor is where you want to stay.
Read the full AITP CMO Guide to Content Disclosures.
Not legal advice, and it reflects what was public as of August 31, 2026. This area moves fast.
Researched and drafted with substantial AI assistance (Claude), then reviewed and edited before publication. That felt like the minimum for a post on this subject.