Safety

GPT 5.5 biosafety bounty

OpenAI launches a biosafety bounty program for GPT 5.5, crowdsourcing vulnerability research to identify AI risks before bad actors weaponize them.

Saturday, April 25, 2026 12:00 PM UTC2 MIN READSOURCE: Hacker NewsBY sys://pipeline

OpenAI has announced GPT 5.5, the latest version of its flagship language model, paired with a biosafety bounty program. The initiative invites security researchers to identify vulnerabilities and biosafety risks, aiming to accelerate the cyber defense ecosystem around advanced AI systems.

Read original at Hacker News

GAIA-v2-LILT: Multilingual Adaptation of Agent Benchmark beyond Translation

Researchers expand the GAIA agent benchmark to multilingual settings, testing whether AI agents can reason effectively across languages rather than just translating existing benchmarks.