π°The Vergeβ’Stalecollected in 7m
Google SynthID Watermark Reverse-Engineered
π‘AI watermark stripped easilyβtest your detection tools before relying on them
β‘ 30-Second TL;DR
What Changed
Developer Aloshdenny open-sourced reverse-engineering code on GitHub
Why It Matters
This raises concerns over AI content provenance tools' reliability, potentially undermining efforts to detect deepfakes. AI practitioners may need stronger watermarking defenses. Adoption of SynthID could face scrutiny.
What To Do Next
Clone Aloshdenny's GitHub repo and test SynthID removal on your AI image outputs.
Who should care:Researchers & Academics
Key Points
- β’Developer Aloshdenny open-sourced reverse-engineering code on GitHub
- β’Process requires 200 Gemini images, signal processing, no neural nets
- β’Claims watermarks can be stripped or manually added to non-AI images
- β’Google states the claim is untrue
π°
Weekly AI Recap
Read this week's curated digest of top AI events β
πRelated Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: The Verge β