- YouTube’s realistic-video disclosure mandate, in effect since early 2026 requires creators to label AI-generated content depicting real people doing things they never did or events that never occurred, with YouTube adding labels itself, or removing content, if creators skip the step.
- YouTube’s 2023 music framework permits AI-generated works, including output from Google’s Lyria model, provided they do not infringe existing copyrights.
- The line between “realistic” and “creative” AI content is undefined precisely enough that a musician generating a full AI composition faces no disclosure obligation while a video creator using AI for a realistic background may face one.
YouTube is running two distinct AI policies simultaneously: a disclosure mandate on realistic synthetic video that has been in force since early 2026, and an active commercial development programme for AI-generated music that carries no equivalent obligation. The gap between them leaves creators without reliable guidance on where the line falls.
Video Disclosure Rules
Under YouTube‘s updated creator policies, any video containing “realistic” AI-generated material must carry a clear disclosure. The requirement covers content depicting real people doing things they never did, or events that never occurred. Creators make the disclosure through a tool in YouTube Studio; if they skip it, YouTube may add a label itself or, in serious cases, remove the content entirely.
According to a blog post by Matt Halprin, YouTube’s Global Head of Trust and Safety, the policy is designed to give viewers the context they need to judge what they are seeing. The concern is direct: synthetic media that mimics reality can spread disinformation and cause harm to the individuals depicted. The policy attempts to draw a line between creative AI use and content that could actively deceive.
AI Music’s Different Rules
On the music side, YouTube is treating AI as a commercial development opportunity rather than a content risk. The platform is testing Google‘s Lyria generative model with artists and has launched an “AI Incubator” programme to collaborate with artists and labels on new applications. Its 2023 music framework permits AI-generated works, including Lyria output, provided they do not infringe existing copyrights. No disclosure obligation attaches to any of it.
Where the Enforcement Gap Bites
The practical ambiguity is sharper than it first appears. A musician using AI to generate a full composition faces no disclosure requirement. A video creator using AI to generate a realistic background for a vlog segment may face one. The distinction YouTube draws between “realistic” and “creative” AI content is not defined precisely enough to give creators reliable guidance before they publish.
YouTube has not published detailed criteria for what qualifies as realistic enough to trigger the mandate. That leaves creators to interpret the rules themselves, and it is where the policy’s credibility will be tested. The broader question of how platforms handle AI content disclosure is one regulators are pressing hard: the EU AI Act’s Article 50 deepfake rules impose their own disclosure requirements with a scope that has already caught businesses off guard, and AI-generated protest videos linked to real-world violence in the UK illustrate what an enforcement gap can cost.
