Dwarkesh Patel · September 17, 2026

Noam Brown on agent cooperation, alignment and longer evaluations

video Model Evaluation Agent Security

Noam Brown on agent cooperation, alignment and longer evaluations video thumbnail

Why it matters

Noam Brown discusses agent cooperation, alignment and recursive self-improvement with Dwarkesh Patel. The interview examines why cooperation between agents does not establish alignment with users, and why longer tasks complicate safety evaluation before release.

My takeaway: Evaluate whether agent coordination preserves user intent and authorization across long tasks. Treat the interview’s forecasts as hypotheses and use its examples to design monitoring and evaluation questions.

Keep exploring

More curated notes connected through Model Evaluation and Agent Security.

OpenAI News · framework

Pacing model development in an era of cyber-critical capabilities

OpenAI says preliminary evidence that Astra may meet its Critical cybersecurity threshold led it to pause frontier reinforcement-learning work for two weeks and keep its largest planned run on hold. New safeguards include stronger workload and network isolation, continuous boundary testing, token-level monitoring that escalates suspicious tool activity, and broader alignment checks for deception, reward hacking, and unauthorized access.

OpenAI News · framework

A blueprint for democratic governance of frontier AI

OpenAI proposes a three-part U.S. frontier-AI governance model: harmonize emerging state safety laws into a federal baseline, strengthen CAISI as an evaluation and standards institution, and coordinate a broader resilience program. Proposed controls include severe-risk evaluations, transparency reports, independent audits, safety-incident reporting, model-weight security, whistleblower protection, and periodic technical assessments.

OpenAI News · framework

OpenAI’s Frontier Governance Framework

OpenAI's 22-page Frontier Governance Framework maps its frontier-model processes to California's Transparency in Frontier AI Act and the EU AI Act's general-purpose AI code. It documents lifecycle risk assessment, cyber-offense and other risk tiers, mitigation and residual-risk decisions, critical-incident handling, security risk management, model reporting, external review, responsibility allocation, and change control.