Emerging AI regulation is moving fast to require machine-readable transparency disclosures on AI-generated content.
By the end of 2026, the EU AI Act's transparency obligations under Article 50 will be fully in effect. Penalties for non-compliance are steep - up to €15 million or 3% of global annual turnover, whichever is higher. Time is short.
The industry is rapidly converging on C2PA as the digital standard for "machine-readable" provenance, and the strongest available standard for conforming with Article 50(2) currently.
The Coalition for Content Provenance and Authenticity (C2PA) has been working on content provenance since 2021. It runs a formal Conformance Program that ensures generator, validator and certificate authority products actually meet their rigorous spec and security requirements.
We’re proud to share that Monolith recently achieved C2PA Generator conformance after a thorough technical conformance exercise with C2PA’s Technical Working Group - and that Monolith is now the easiest way to comply with the EU AI Act's Article 50(2) machine-readable disclosure requirements.
Why we built Monolith
We originally built Monolith to help the media industry record copyright claims and distinguish human vs AI provenance. We pursued C2PA conformance so Monolith would be the easiest way for anyone to adopt the C2PA standard - to enhance everyone's provenance capabilities.
At the time, we didn't realize that achieving conformance would also make Monolith the easiest way to comply with the EU AI Act's Article 50(2) and other emerging AI regulations demanding machine-readable transparency disclosures, but here we are - insight is always sharpest in hindsight.
How Monolith meets the EU AI Act Article 50(2) requirements
Briefly, Article 50(2) stipulates that Providers of AI systems need to include (i) machine-readable marks indicating if a (ii) video, audio, image, or text file was (iii) AI-generated or AI-manipulated.
(i) Monolith records are machine-readable
Monolith records are machine-readable by design. Additionally, Monolith has achieved C2PA Generator conformance, meaning Monolith records are also machine-readable by any C2PA verifier. C2PA is the current de-facto digital standard for metadata relating to content provenance and authenticity.
You can easily include machine-readable marks indicating if a file was AI-generated or AI-manipulated using the Monolith Enterprise tier.
(ii) Monolith supports the required file types…and more
Monolith has achieved C2PA conformance for the most number file types ever granted by C2PA to a single generator. But Monolith is designed to generate metadata for all file types - not just C2PA-recognized ones.
So you can record machine-readable AI-generated or AI-manipulated marks for all video, audio, image, or text files, including any unique or emerging filetypes now or in the future.
(iii) Monolith lets you mark files as AI-generated or AI-manipulated
The Monolith Enterprise tier allows you to embed custom metadata, including marking files as AI-generated or AI-manipulated.
For the purposes of EU AI Act Article 50(2), a simple AI-generated or AI-manipulated mark will suffice, but in practice, enterprises often want to include additional metadata such as copyright assertions, production data, reference or serial numbers, and more.
Monolith can support all custom metadata needs, including the machine-readable transparency disclosures required under the EU AI Act.
Monolith vs other C2PA Generators
Monolith offers a few key advantages over other C2PA Generators.
- Universal filetype support: Most other C2PA Generators only support specific filetypes. Monolith’s universal filetype support means you just need to use a single platform to handle all filetypes.
- Single pricing structure: If you use multiple C2PA Generators, you will incur different pricing rates. Monolith offers a single, transparent, and affordable pricing structure to enable you to mark files at scale.
- Ease of integration: Monolith Enterprise is designed to be extremely easy to integrate. So you don’t need to juggle different platform interfaces for different filetypes - you can simply integrate Monolith Enterprise into your existing systems using our API. From that point on, all files you designate on your existing systems will get marked and become verifiable as AI-generated or AI-manipulated (along with whatever other custom metadata you require) through Monolith Enterprise.
- Robustness and proving fulfilment of duties: Monolith stores copies of the metadata separately from the files (called a “sidecar manifest”). These sidecar manifests are associated with your file using your file’s unique fingerprint (called a “hash”). If your AI-generated or AI-manipulated mark gets stripped from your file by a metadata eraser (and there are so many free tools for this online) you can easily prove that you fulfilled your duties at the relevant time by marking the file as AI-generated or AI-manipulated at that time. Sidecar manifests are stored on a publicly available network for robustness and open verifiability.
Speed, scale, simplicity, filetype compatibility, affordability, robustness, and public verifiability - these all converge in Monolith to make it the easiest way for enterprises to comply with machine-readable transparency disclosures.
If you’re a Provider of an AI system under the EU AI Act, get in touch so we can support your Article 50(2) compliance efforts.