From 2 August 2026, providers of generative AI systems in the European Union must mark their outputs in a machine-readable way and make them detectable as AI-generated. Systems already on the market before that date have until 2 December 2026.
ORIproof gives you a ready-to-use labelling setup, combining C2PA Content Credentials and imperceptible watermarking, so you can comply in time and keep shipping.
For AI platforms, creative agencies, publishers, and organizations producing AI-generated images and video, transparency is becoming an operational requirement. ORIproof helps bring provenance into your content workflow – so information about AI use can be recorded, checked, and recovered as media moves between platforms.
A global shift towards AI transparency
Regulators are introducing requirements to identify AI-generated content, preserve provenance, and inform audiences. As of this writing, the details differ by market:
| Region | Law | Applies from |
|---|---|---|
| European Union | AI Act, Art. 50: machine-readable marking and detection of AI content, deep-fake disclosure | 2 Aug 2026 (existing systems: 2 Dec 2026) |
| China | AI Labelling Measures + GB 45438-2025: visible labels plus metadata labels | 1 Sep 2025 |
| South Korea | AI Basic Act: label generative-AI outputs | 22 Jan 2026 (fines after a grace period) |
| California (US) | AI Transparency Act (SB 942 / AB 853): embedded, machine-readable provenance disclosures and a free detection tool | 2 Aug 2026, platforms 1 Jan 2027, devices 1 Jan 2028 |
| India | IT Rules amendment: prominent labels plus embedded metadata | 20 Feb 2026 |
| Vietnam | AI Law: machine-readable marks on AI audio, images, and video | 1 Mar 2026 |
The EU AI Act, briefly
- Providers must mark AI-generated audio, images, video, and text in a machine-readable way and make them detectable. Marking and detection are separate duties.
- Marking must be effective, interoperable, robust, and reliable as far as technically feasible. The EU Code of Practice expects signed metadata plus an imperceptible watermark.
- Deployers must disclose deep fakes and certain public-interest text, with a lighter disclosure for creative or satirical works.
- Fines reach €15 million or 3% of worldwide turnover. The company offering the AI system under its own name is responsible, even if it uses a third-party model.
How ORIproof helps
Record AI provenance with C2PA
Build a workflow around signed Content Credentials that document declared AI generation or editing and relevant production history in a standard, verifiable format.
Keep a recoverable connection to that provenance
Imperceptible watermarking embeds an identifier into images and video frames, linking media to its associated provenance record. This supports recovery when embedded metadata is lost during sharing or processing – the approach C2PA calls soft binding. C2PA guidance
Make transparency part of publishing
Connect provenance marking and verification with the audience-facing disclosures your publishing workflow requires. Machine-readable credentials and visible labels serve complementary purposes.
Offer free detection, as the law expects
A public verification page lets anyone check whether content carries your AI label, and shows whether the result comes from the metadata or the watermark. The EU AI Act Article 50(2) requires outputs to be detectable, not only marked.
Run it where your content lives
A ready-to-run signer on your own infrastructure, or an EU-hosted API. Your content never has to leave your systems, and there’s no dependency on a service outside the EU. This is valuable if your organization has technology sovereignty requirements.
Trusted EU certificates and timestamps
Signing certificates and trusted timestamps from Castlabs, a Berlin-based certificate authority whose root is on the official C2PA trust lists. We keep the signer and certificates current as standards evolve
Make your AI content easier to identify – and its provenance easier to verify
Get ready before 2 December 2026: start a free trial, or talk to us about going live in your own environment or ours.
ORIproof supports the technical implementation of AI transparency measures. Compliance depends on your role, use case, jurisdiction and implementation – C2PA credentials alone don’t establish compliance with every applicable obligation.
Regulatory information reviewed: 09 November 2026
Recent updates
RePro Stream & STARDUSTmark: Bringing forensic accountability to remote film and TV production
Challenge: Securing distributed production Modern productions rely on teams who are rarely in the same...
DRM and the modern content security stack – IBC 2026
How do different content security layers work together to protect premium video? Watch the full...
IAMT Studio interview with Christophe Kind at IBC 2026
How can media companies protect their streams, grow advertising revenue, and help audiences trust what...
