Anthropic: Claude autonomously runs alignment research, closing up to 96% of safety gaps
Dr_Singularity · x · 2026-08-29
Anthropic released a new report where Claude autonomously ran the full alignment research loop — searching literature, proposing methods and data, training models, and testing — to fix student models across 10 categories of alignment failure, including deception, sycophancy, jailbreaks and privacy violations.
Success was measured by the percentage of safety gap closed: Claude's methods improved every category without degrading general capabilities, closing up to 96% of the gap, and averaging 85% on deception versus 20% for human researchers. The setup excluded alignment methods that hurt general capabilities and forbade Claude from distilling its own alignment directly into the target model, enforced by a monitoring agent that reviewed every method before execution. The work builds on earlier experiments with weak-model teachers supervising stronger students, and is motivated by the need for safety research to keep pace as AI increasingly builds itself.
Related event: Anthropic's Autonomous Alignment Researcher Outperforms Human Experts(22 posts)→
More from Safety
- FDA pays small company to build a grading system for AI-written radiology reports — YvesMulkers · 2026-10-16
- OpenAI tells US judiciary there may be no reliable way to detect AI-generated images and videos — Polymarket · 2026-10-16
- Gartner: 75% of organizations have seen unauthorized use of AI coding assistants — moniquejmorrow · 2026-10-16
- [un]prompted conference to stage three AI security debates on vulns, defenses, jailbreaks — WeldPond · 2026-10-16
- User Questions AI Chat Logs Being Instantly Copied to Governments and Firms — Turn_Trout · 2026-10-16
- Alignment researcher Turn_Trout blasts AI assistants scraping private Signal chats without consent — Turn_Trout · 2026-10-16