Markets US

Over 100 AI experts, including Geoffrey Hinton and members of Stanford University and Johns Hopkins University, have signed a public letter urging Anthropic and OpenAI to adopt independent third-party safety evaluators. The letter, shared exclusively with CNBC, argues that current safeguards are insufficient to assess AI risks, calling for "scientific objectivity, transparency, independence, and robust protections" for evaluators. The demand reflects growing industry pressure to standardize AI risk management practices amid concerns over potential security vulnerabilities.

The letter, published on September 18, 2026, calls for foundation model providers to grant third-party evaluators "employee-like access" to inspect and audit their development processes. Anthropic CEO Dario Amodei proposed the idea, which has been endorsed by OpenAI CEO Sam Altman, SpaceX's Elon Musk, and Microsoft CEO Satya Nadella. The signatories include the nonprofit evaluator METR and academic institutions, emphasizing the need for evaluators to have full editorial control and be shielded from corporate retaliation.

The coalition advocates for standardized evaluation frameworks, including the AEF-1 standard, to ensure credible oversight. Evaluators would need access to sensitive data, candid employee communications, and unreleased systems to assess risks effectively. However, logistical challenges remain, such as determining which evaluators to select and how deeply they can inspect proprietary technologies. The letter warns that without independent oversight, companies risk undermining public trust and national security, particularly as a few labs control critical AI capabilities.

Stakeholders like Anthropic and OpenAI face heightened scrutiny over their risk management practices. Regulators may need to establish oversight frameworks if companies fail to comply. Investors could see market shifts based on corporate responses, while the public relies on companies to ensure AI safety. The letter highlights tensions between corporate control and external oversight, with implications for regulatory scrutiny and investor confidence.

The letter underscores the urgent need for independent AI safety protocols, but its effectiveness hinges on corporate compliance and regulatory action—areas that remain unresolved. Without standardized oversight, the risk of unaddressed vulnerabilities in frontier models persists, raising questions about how companies will balance innovation with public safety.

Topics: AI Safety, Technology Regulation, Corporate Governance, Artificial Intelligence, Risk Management, Third-Party Oversight, AI Ethics, AI Governance Standards, AI Security, Innovation Oversight

#AIRegulation #TechGovernance #AI Safety #CorporateOversight #AIStandards #AI Ethics #InnovationControl #TechSecurity

Source: CNBC