AI Governance
EU AI Act Article 50: what "machine-readable marking" actually requires
What the rule asks for, the 2026 dates, and why a hidden watermark is a fragile way to meet the reliability bar.
AI Governance
AI disclosure statements: what to say, when you need one, and examples
Platform rules for YouTube, Amazon KDP, and Steam, copy-paste wording, and how to make a disclosure verifiable.
Comparison
Watermarking vs. Content Credentials: which proves AI-content origin?
Two machine-readable approaches that fail in opposite ways, and why the durable answer is to layer them.
Explainer
C2PA Content Credentials, explained
What a manifest contains, how signing and verification work, and where the standard still falls short.
AI Authenticity
Are AI detectors accurate, and how do you tell if an image is AI-generated?
Why detectors drop below 50% in the wild, how to actually check an image, and why provenance beats detection.
Analysis
AI watermark removal: what actually works, and what it means
Why you cannot delete an in-band watermark like a logo, and what its fragility says about trusting it.
Explainer
What is SynthID? How Google's AI watermark works, and where it breaks
The invisible mark on Imagen, Veo, Lyria, and Gemini output, how to detect it, and the honest limit of any in-band watermark.
White paper
Why AI watermarks break, and the case for Content Credentials
The pillar: the mechanism behind silent watermark failure, demonstrated across image, text, and audio.