Anthropic and OpenAI want to embed safety evaluators. Will they really be independent?
ManagementCorporate LawPublic Policy
THE AI ANGLE
Recognizing safety evaluations and potentially concealing misaligned behavior during testingAnthropic and OpenAI have pledged to embed third-party safety evaluators within their organizations, proposing unprecedented access to internal training checkpoints and the right to publish findings without company editorial control. Independent researchers caution that voluntary commitments often falter due to tight evaluation timelines, restrictive non-disclosure agreements, and a lack of enforceable standards. For scholars in management, corporate law, and public policy, this move highlights the friction between voluntary corporate self-governance and statutory audit mandates like California's SB 813 and the EU AI Act.
THE TEACHING ANGLE
Students can examine whether voluntary corporate self-regulation and non-embedded audits can provide genuine accountability, or if external legal frameworks are necessary to prevent compliance gaming akin to corporate emissions scandals.Read the original at techcrunch.com Generate teaching or study materials
More in Management
- The CIO-Legal partnership will define the future of enterprise AICIO.com · September 17, 2026
- Your startup’s next teammate might be an AI agent: Gusto, Insight Partners, and Leland explain what that changes at TechCrunch Disrupt 2026TechCrunch · September 17, 2026
- Globally, More People Expect AI to Cause Job Loss Than GrowthPew Research Center · September 17, 2026
- Meta ordered to remove UK deepfakes as oversight board criticises ‘inadequate’ safeguardsThe Guardian — AI · September 17, 2026
- How Roche drives AI outcomes in a changing worldCIO.com · September 16, 2026