Theories of Change for AI Auditing

Apollo Research's framework for how AI auditing reduces risk from misaligned AI.

Executive Summary Our mission at Apollo Research is to reduce catastrophic risks from AI by auditing advanced AI systems for misalignment and dangerous capabilities, with an initial focus on deceptive alignment. In our announcement post, we presented a brief theory of change of our organization which explains why we expect AI auditing to be strongly positive for reducing catastrophic risk from advanced AI systems. In this post, we present a theory of change for how AI auditing could improve the safety of advanced AI systems. We describe what AI auditing organizations would do; why we…

saved by

• N . Soma Sekhar

related reading

• Recommendations for Technical AI Safety Research Directionsalignment.anthropic.com
• Auditing the singularitywuthejeff.com
• The case for ensuring that powerful AIs are controlled — LessWronglesswrong.com
• Auditing language models for hidden objectives \ Anthropicanthropic.com
• Auditing language models for hidden objectives — LessWronglesswrong.com
• [2408.02565] Reasons to Doubt the Impact of AI Risk Evaluationsarxiv.org
• AI Control: Improving Safety Despite Intentional Subversion — LessWronglesswrong.com
• AI Control: Improving Safety Despite Intentional Subversion — AI Alignment Forumalignmentforum.org
• We need 3rd party Training-Run Assessments — LessWronglesswrong.com
• AI models can be dangerous before public deploymentmetr.org
• [2004.07213] Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claimsarxiv.org
• Auditing failures vs concentrated failures — AI Alignment Forumalignmentforum.org