GENERAITR Compliance
Automatic EU AI Act compliance for generative pipelines — every asset perceptually hashed, invisibly watermarked, and C2PA-signed the moment it's generated.
The problem
The EU AI Act requires every organization deploying generative AI to mark synthetic content in a machine-readable format (Article 50), with penalties up to €15M or 3% of global turnover per breach. Bolting compliance on after the fact means retrofitting watermarking and audit trails into a production pipeline that was never built to carry them — expensive, and easy to get wrong quietly.
What it does
GENERAITR Compliance makes provenance automatic at generation time, not a separate step someone has to remember to run. Every generated asset is registered with a sequential ID, fingerprinted with a perceptual hash that survives resizing and re-compression, invisibly watermarked (pixel-level for images, per-frame for video with the original audio reconstructed), and signed with a C2PA manifest — the same Content Credentials standard the EU AI Act points to as its reference implementation.
Why trust it
Built around three pillars: systematic (an auditable lineage database, not tribal knowledge), compliant (every generation documented, always audit-ready), and responsible (on-premise deployment option, GDPR-aligned storage — the data stays the customer's). None of it depends on an end user remembering to opt in.
Current maturity
Running in production today as part of GENERAITR SaaS's own generation pipeline — every asset produced there already carries this. Packaging it as an independently callable standalone module — the "lösvikt" offering for pipelines that don't run on GENERAITR at all — is the next phase of work.