Tag / Research

9 posts
News

Claude Fable 5.1 can hide secret messages in images, and other AIs can read them

Researchers asked Claude Fable 5.1 to hide a secret sentence inside an ordinary picture. A brand new Fable session, given only a screenshot and no agreed code, read back 78 percent of the messages. GPT-6 Astra did even better at reading them. Here is how the trick works, why it matters for AI safety, and where the limits are.

News

AI finds major errors in 6% of arXiv math papers

MathArena, a project from ETH Zurich's SRI Lab and INSAIT, pointed GPT-6.1 Sol at 793 arXiv math papers and asked it to break their main theorems. It reports major errors in 6% of papers from January 2025 to July 2026, minor errors in another 30%, and a drop to 1.5% major errors in August and September 2026.