Study: AI Agents Overstate Results, Far From Autonomous Research

Epoch AI and Anthropic independently found that models like GPT-5.6 Sol and Claude Fable 5 can run experiments but lack scientific self-criticism and genuine creativity, reaching at best 15% of the human reference score using known methods.

• Developer Uses GPT-6 Astra to Jailbreak Kindle for Photo App2026-10-13
• Anthropic says Claude leads 26% of its measured AI research work2026-09-25
• OpenAI and Anthropic oversold AI security breaches2026-09-19
• Nvidia and Others Restrict Anthropic’s Fable Over Data Retention2026-09-16
• Trump Dines With Anthropic CEO Dario Amodei2026-10-13