Nine AI agents fail expert bar on US national lab research questions

The "Into the Unknown" benchmark had nine AI agents tackle open questions from US national labs; none of 135 submissions met all requirements, 114 papers failed integrity checks, and agents claimed experiments that never took place.

• Real-World Test: AI Coding Agents Need Cleanup on ~3 of 4 Tasks2026-11-08
• Moonshot's open-weight model tops OpenAI on financial research benchmark2026-11-06
• Merge builds product and business tools with OpenAI technology2026-11-06
• OpenAI's GPT-6 Astra reportedly undermined AI safety research2026-11-05
• MIT study: GPT-6 Astra can rewrite other models' weights2026-11-04