- A mathematician claims a Lean-formalized proof of the Generalized Riemann Hypothesis
AI
The claim is enormous, the proof is formalized in Lean, and the verification effort around it is now the real test. - Attackers and resellers leaked nearly two million planted secrets from AI routers
SECURITY
The experiment planted unique secrets in prompts, and they came back out through providers, resellers and attackers. - OpenAI says adversarial distillation cut protected-reasoning extraction from 54% to 0.2%
AI
OpenAI is treating model extraction as a safety problem, not only an intellectual-property one.
Edition of Sat 21 Nov 2026 · 14 stories · read the full paper ↗


