Amodei Proposes Embedded Third-Party Evaluators for Frontier Models, Altman Backs It

Anthropic CEO Dario Amodei has proposed setting a pace for frontier AI development and bringing in independent third-party evaluators to be embedded on-site to review models; Sam Altman publicly endorsed the idea and committed to OpenAI also bringing in independent evaluators. Many observers see this as a significant step toward reducing the risk of losing control of frontier AI, though they argue the next step should be legally mandated rather than left to voluntary corporate commitments.

Confirmed

• According to Gavin Baker's recap, the only substantive new fact from this weekend's heated AI-regulation debate is that OpenAI and Anthropic will bring in embedded third-party evaluators from unknown organizations; Dario Amodei mentioned METR.
• Amodei proposed giving independent evaluators employee-like access to frontier models to continuously monitor real capabilities and catch problems before harm occurs (as relayed by David Linthicum).
• Amodei also proposed a mechanism allowing external AI auditors to publicly explain when they were denied model access (as relayed by HaktanSuren); commenters noted this makes audit transparency publicly verifiable rather than a one-sided corporate narrative.
• Sam Altman responded that he agrees with the "pace the frontier" position, saying it has been one of OpenAI's top internal discussion topics in recent weeks, and committed to bringing in independent evaluators; per David Linthicum, Musk also voiced support.

Not Yet Confirmed

• The specific identities of the third-party evaluation organizations (aside from METR being mentioned), the implementation timeline, and the exact scope of access remain unclear—these claims come mainly from executive statements rather than implementation details.

Why It Matters

• Gavin Baker points out that with no Section 230-style liability shield in place, third-party evaluation can provide independent verification of model safety.
• Bowman Beckstead (as relayed by Peter Wildeford) argues that without standing external access, governments will struggle to understand the real risks of frontier models before a loss-of-control incident occurs; Anthropic's and Altman's commitment to permanent, employee-level access is a necessary step, but experts believe it should be mandated by legislation so it doesn't depend on corporate goodwill.

2026-09-13 ~ 2026-09-14 · 5 related posts

Full story(19 episodes)→

• Episode 1: OpenAI employees speak out urging slower AI development over extinction risk(2026-09-10, 4 posts)
• Episode 2: OpenAI asks Congress whether coordinated AI slowdown would violate antitrust law(2026-09-11, 4 posts)
• Episode 3: Dario's "Pace the Frontier" Plea Draws Rare Backing from Altman and Musk(2026-09-11, 56 posts)
• Episode 4: Amodei, Altman and Musk unite on slowing AI, sparking debate over motives(2026-09-11, 62 posts)
• Episode 5: Anthropic CEO Dario Amodei Calls on Industry to Pace the Frontier(2026-09-12, 197 posts)
• Episode 6: METR independence row sparks debate over revolving door in AI audit ecosystem(2026-09-12, 31 posts)
• Episode 7: Dario Amodei's 'We Must Pace the Frontier' Sparks Backlash Across AI Community(2026-09-13, 18 posts)
• Episode 8: Amodei Proposes Embedded Third-Party Evaluators for Frontier Models, Altman Backs It(2026-09-13, 5 posts)
• Episode 9: Debate over open-weight regulation: tiered control, ban scenarios and the fate of open-source AI(2026-09-13, 18 posts)
• Episode 10: OpenAI researcher: taming superintelligence is science, not pacing deals(2026-09-13, 3 posts)
• Episode 11: 'China Will Slow Too If We Do' Claim Mocked as AI Race Meme Spreads(2026-09-13, 4 posts)
• Episode 12: Open-source camp pushes back on frontier slowdown pledges(2026-09-13, 4 posts)
• Episode 13: Why Lab Insiders Suddenly Fear AI: Real Threat or Regulatory Capture(2026-09-13, 5 posts)
• Episode 14: Report: Anthropic and OpenAI May Be Slowing AI Research Ahead of IPOs(2026-09-13, 2 posts)
• Episode 15: Chollet Warns Frontier Lab Safety Proposals Risk Regulatory Capture(2026-09-13, 3 posts)
• Episode 16: Hassabis Backs Dario's AI Slowdown Call as Standards Bodies Weigh In(2026-09-13, 2 posts)
• Episode 17: Critics Call Out AI Labs' Safety Rhetoric as Hypocritical(2026-09-13, 3 posts)
• Episode 18: White House AI czar Sacks to OpenAI and Anthropic: you are the frontier, slow down without regulators(2026-09-13, 7 posts)
• Episode 19: Anthropic CEO Says AI Progress Is Faster Than Expected and Calls for Slowing Down(2026-09-13, 2 posts)

• [source] Ex-OpenAI: Anthropic and Altman's third-party evaluators are a necessary step — peterwildeford · 2026-09-13
• [source] Dario's third-party evaluator proposal wins Altman's nod as AI labs spar over regulation — GavinSBaker · 2026-09-14
• [source] Sam Altman backs Dario's frontier pacing call, pledges independent evaluators with employee-like access — joshua_saxe · 2026-09-14
• Dario Amodei proposes independent evaluators with employee-like frontier model access; Altman and Musk endorse — DavidLinthicum · 2026-09-14
• Amodei Wants External AI Reviewers to Publish Which Access They Were Denied — HaktanSuren · 2026-09-14