Wire · founder news, decoded · technology
Meta AI image detector fails to identify some of its own cropped AI images, Reuters analysis finds
Meta's AI image detector fails to identify 55% of its own cropped images, according to Reuters testing. The watermark-based Content Seal system loses its signal after common edits like cropping, limiting its ability to flag deepfakes ahead of the 2026 US midterms.
This Wire brief sits within Fusion42's coverage of AI Infrastructure and Generative AI. Wire is Fusion42's founder-focused intelligence feed: each story is connected to the funds and startups it names — every one with a live profile on Raise or Scout — so founders can follow the capital and the momentum behind the headline rather than just the headline itself. Wire analysis is one of the live surfaces Arthur, Fusion42's AI co-founder, reasons over.
The Wire takeaway
Meta's detection tool fails on cropped images, which means any platform relying on invisible watermarks for deepfake moderation has a known gap you can exploit. Build detection that works on edited images, or build the attack that proves watermarking is theatre.
Read the full story at reuters.com →
Topics: AI Infrastructure · Generative AI · watermarking-limits · deepfake-detection · content-verification · election-integrity