{"version":1,"type":"story","url":"https://digestai.news/story/ai-labs-turn-to-independent-evaluators-as-safety-concerns-grow","json":"https://digestai.news/story/ai-labs-turn-to-independent-evaluators-as-safety-concerns-grow.json","markdown":"https://digestai.news/story/ai-labs-turn-to-independent-evaluators-as-safety-concerns-grow.md","slug":"ai-labs-turn-to-independent-evaluators-as-safety-concerns-grow","headline":"AI labs turn to independent evaluators as safety concerns grow","summary":"Anthropic and OpenAI are embedding third-party evaluators like METR, Apollo Research, and Transluce to assess model risks, following pressure from researchers and a voluntary White House accord signed by top tech executives. Anthropic CEO Dario Amodei pledged to embed evaluators with access comparable to internal teams, and OpenAI said it is finalizing contracts with groups including METR and Redwood Research. The evaluators, mostly small nonprofits, are seeing a surge in funding: METR raised about $71 million in six months, up from $13.6 million in all of 2024, while for-profit Vals AI grew to 30 employees and announced a $40 million round in August.\n\nFriction is already emerging. OpenAI fired three employees last week, including Mikita Balesni and Tomek Korbak, who said they were dismissed over communications with third-party evaluators. OpenAI disputed this and said it remains committed to external assessments. Critics warn that without independent funding and government standards, evaluators risk becoming beholden to the labs they audit. The FRONTIER Act in Congress and new California laws aim to create licensed Independent Verification Organizations, but federal action remains uncertain. Anthropic and OpenAI have endorsed the California bills while preferring federal requirements.","keyPoints":["Anthropic and OpenAI are embedding independent evaluators like METR and Apollo Research to assess AI model safety","METR raised about $71 million in six months; Vals AI announced a $40 million funding round in August","OpenAI fired three employees, including two who said they were dismissed over contacts with third-party evaluators"],"whyItMatters":"As AI labs race toward trillion-dollar valuations, independent evaluation is becoming the primary safety check absent federal regulation, but funding and access dependencies raise conflict-of-interest concerns.","category":{"slug":"policy","name":"Policy & Regulation","url":"https://digestai.news/category/policy"},"entities":{"companies":["Anthropic","OpenAI","METR","Apollo Research","Transluce","Accenture"],"models":[],"people":["Dario Amodei","Sam Altman","Donald Trump","Suresh Venkatasubramanian","Mikita Balesni","Tomek Korbak"]},"firstPublishedAt":"2026-10-11T14:08:44Z","updatedAt":"2026-10-11T14:08:44Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"CNBC Technology","title":"AI’s quiet safety gatekeepers are stepping into the spotlight","url":"https://cnbc.com/2026/10/11/ais-quiet-safety-gatekeepers-are-stepping-into-the-spotlight.html","publishedAt":"2026-10-11T14:08:44Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"The Unraveling of Human Control Over AI","url":"https://digestai.news/thread/controlai-calls-for-ban-on-superintelligence-development-amid-rising-safety","storyCount":10},"cite":{"text":"Digest AI, \"AI labs turn to independent evaluators as safety concerns grow\", 11 October 2026, https://digestai.news/story/ai-labs-turn-to-independent-evaluators-as-safety-concerns-grow","publisher":"Digest AI","title":"AI labs turn to independent evaluators as safety concerns grow","datePublished":"2026-10-11T14:08:44Z","url":"https://digestai.news/story/ai-labs-turn-to-independent-evaluators-as-safety-concerns-grow"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}