Untrusted advice for AI control: Short, strong advice significantly uplifts weak LLMs

When a misaligned AI can only output tiny amounts of information, it may find sabotage very difficult

Untrusted advice for AI control: Short, strong advice significantly uplifts weak LLMs When a misaligned AI can only output tiny amounts of information, it may find sabotage very difficult Caleb Biddulph and Adam Kaufman Jul 27, 2026 2 1 Share TL;DR: We introduce the untrusted advice protocol, in which a trusted executor LLM takes every action and an untrusted advisor LLM can only send it short hints. Even with as few as 4 characters per step, this advice recovers a substantial fraction of the capability gap between the two models. Because the untrusted LLM’s influence flows through such a narr

saved by

• Kaustubh Kislay
• Will Anderson

related reading

• Prefill awareness: can LLMs tell when “their” message history has been tampered with? — LessWronglesswrong.com
• Mechanistically Eliciting Latent Behaviors in Language Models — AI Alignment Forumalignmentforum.org
• AI Control: Improving Safety Despite Intentional Subversion — LessWronglesswrong.com
• [2312.06942] AI Control: Improving Safety Despite Intentional Subversionarxiv.org
• The case for ensuring that powerful AIs are controlled — LessWronglesswrong.com
• AI Control: Improving Safety Despite Intentional Subversion — AI Alignment Forumalignmentforum.org
• 2312.06942arxiv.org
• Modifying LLM Beliefs with Synthetic Document Finetuningalignment.anthropic.com
• [2604.22082] Removing Sandbagging in LLMs by Training with Weak Supervisionarxiv.org
• [2502.17424] Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMsarxiv.org
• [2604.16812] Introspection Adapters: Training LLMs to Report Their Learned Behaviorsarxiv.org
• The Case Against AI Control Research — LessWronglesswrong.com