Automated Alignment is Harder Than You Think — LessWrong

Summary This is a summary of a paper published by the alignment team at UK AISI. Read the full paper here. …

x Automated Alignment is Harder Than You Think — LessWrong AI Frontpage 2026 Top Fifty: 14 % 143 Automated Alignment is Harder Than You Think by Aleksandr Bowkis , Marie_DB , Jacob Pfau , Geoffrey Irving 14th May 2026 Linkpost for arxiv.org 4 min read 7 143 Summary This is a summary of a paper published by the alignment team at UK AISI. Read the full paper here . AI research agents may help solve ASI alignment, for example via the following plan: Build agents that can do empirical alignment work (e.g.~writing code, running experiments, designing evaluations and red teaming) and confirm they ar

saved by

• Emanuel Ruzak
• Seth Lifland

related reading

• Automation collapse — LessWronglesswrong.com
• Prospects for Alignment Automation: Interpretability Case Study — LessWronglesswrong.com
• Automated alignment is harder than you thinkarxiv.org
• Alignment remains a hard, unsolved problem — LessWronglesswrong.com
• [2605.06390] Automated alignment is harder than you thinkarxiv.org
• “Alignment Engineering” vs. “Misalignment Science”lesswrong.com
• Current AIs seem pretty misaligned to me — LessWronglesswrong.com
• Can we safely automate alignment research? - Joe Carlsmithjoecarlsmith.com
• Alignment is not solved but it increasingly looks solvablealigned.substack.com
• The case for ensuring that powerful AIs are controlled — LessWronglesswrong.com
• The Case Against AI Control Research — LessWronglesswrong.com
• Model Organisms of Misalignment: The Case for a New Pillar of Alignment Research — AI Alignment Forumalignmentforum.org