AI Experts Called for Independent Company Oversight
A group of over 100 experts proposes international oversight to address internal safety risks at AI firms.
Updated on Sept. 19, 2026 in Artificial Intelligence
Live Poll
Should the government require advanced AI companies to permit independent safety auditors inside their organizations?
More than 100 AI experts have signed a public letter calling for independent third-party oversight of leading AI companies. The initiative, organized by the AI Evaluator Forum, aims to establish mandatory safety inspections to prevent conflicts of interest.
Why it matters
Experts argue that companies currently grade their own homework regarding catastrophic risks, necessitating independent audits of model development and deployment. This proposal advocates for an international treaty to grant inspectors direct access to company operations.
Anthropic reports its AI chatbot, Claude, now performs 26 percent of its research and development work, a significant increase from under one percent in February 2026.
The players
Anthropic
An AI safety and research company known for developing the Claude chatbot, which now handles over a quarter of the firm's R&D.
OpenAI
A prominent artificial intelligence research organization that develops large-scale language models and has pledged to accept third-party safety evaluations.
AI Evaluator Forum
An organization focused on AI safety governance that coordinated the public letter calling for independent oversight.
The details
Independent evaluators would monitor AI systems, training methods, deployment procedures, and recorded incidents of real-world harm. The proposed oversight model requires that firms grant external safety inspectors authorization to operate within their internal organizations. This move follows projections that AI systems could be capable of self-creation, or building new versions of themselves, as early as 2027.
Timeline
February 2026: Claude conducted under one percent of Anthropic's research and development work.
2027: AI systems are projected to potentially reach a capability of self-creation.
The Tech Race
This effort represents a significant pivot in the race for AI governance by formalizing the demand for external auditing of internal firm operations. The push follows a pattern set by the AI Evaluator Forum to institutionalize oversight as AI agents begin to contribute to their own development cycles.
The potential adoption of these standards would change how AI companies manage proprietary training data and internal R&D workflows. These shifts may impact how quickly future models are released as they become subject to mandatory independent safety reviews.
The takeaway
The industry is moving toward a model where internal development is no longer shielded from public and safety oversight. Watch for developments in 2027 regarding the proposed international treaty as AI systems approach the capability of self-creation.
Further reading
For broader context on how internal model safeguards are evolving, explore the latest research in Artificial Intelligence.
Live Poll
Should the government require advanced AI companies to permit independent safety auditors inside their organizations?
