Active means the regulation is in force and enforceable today. Partial means some provisions are active but full enforcement is pending. Proposed means legislation has been introduced but not yet enacted. None means no specific AI content labelling mandate exists in that jurisdiction.
This table covers content labelling and transparency requirements specifically. Most jurisdictions also have broader AI governance frameworks that address other issues (safety, bias, accountability) not covered here.
Key takeaways
The EU is the global benchmark. The AI Act's Article 50 is the most comprehensive, specific, and enforceable content labelling mandate in the world. Other jurisdictions are watching its implementation closely and many will use it as a template.
China moved first on AI content labelling, with the Deep Synthesis Provisions taking effect in January 2023 - over two years before the EU AI Act's transparency obligations became enforceable. China's approach is narrower in scope but was the first binding regulation globally to mandate labelling of AI-generated content.
The US remains fragmented. No comprehensive federal legislation has been enacted. The executive order provides direction but lacks enforcement teeth. State-level action is creating a patchwork of requirements that varies by jurisdiction and context (particularly elections). This fragmentation actually strengthens the case for adopting an open standard like C2PA - a single technical implementation that satisfies multiple jurisdictions.
No jurisdiction mandates C2PA specifically, but the EU's requirement for "machine-readable, interoperable, robust" labelling maps directly to what C2PA provides. As harmonised standards are developed under the AI Act, C2PA (or something very close to it) is likely to be formally recognised as a satisfying mechanism.
Extraterritorial reach is the norm. The EU AI Act, China's regulations, and most proposed US legislation apply to content that reaches their jurisdiction, regardless of where the provider is based. This means a global company effectively needs to comply with the strictest applicable regulation - which is currently the EU.
What this means for organisations
If your organisation generates AI content that reaches users in multiple jurisdictions - which describes most technology companies - the practical approach is to implement the highest standard and apply it globally. Currently, that means satisfying the EU AI Act's Article 50 requirements, which in practice means attaching C2PA Content Credentials to AI-generated outputs.
This approach has three advantages. First, it satisfies the most stringent current requirement. Second, it positions you well for regulations that are still being developed in other jurisdictions (which are likely to follow the EU's lead). Third, it's a single technical implementation that you can apply uniformly, rather than a patchwork of jurisdiction-specific solutions.
For a detailed walkthrough of Article 50 compliance, see our EU AI Act Compliance Guide. For the broader policy context, see The Policy Landscape.
This comparison is maintained by the C2PA.ai editorial team and does not constitute legal advice. Regulatory landscapes change frequently - verify current requirements with qualified legal counsel. Last updated March 2026. Contact us with updates or corrections.
Related: EU AI Act Compliance Guide · The Policy Landscape · Consulting Services