• Home
• News

Anthropic and OpenAI Move Toward Independent Oversight of AI Systems

Anthropic and OpenAI are moving toward greater involvement of independent evaluators in monitoring the safety of advanced artificial intelligence systems.

The initiative follows calls for external experts to gain broader and more continuous access to AI development environments, allowing them to assess models during development and identify potential risks before they reach users.

Anthropic CEO Dario Amodei has advocated for independent evaluators to be given access to AI laboratories, arguing that external oversight could help identify unexpected behaviors and emerging safety risks at an earlier stage.

OpenAI has also expressed support for the concept. CEO Sam Altman indicated that the company is open to adopting a similar approach to independent AI evaluation.

Unlike traditional safety assessments, which often take place shortly before or after a model is completed, continuous external monitoring could allow researchers to observe how AI systems behave throughout different stages of development.

However, several details remain unresolved, including which organizations would conduct the evaluations, what level of access they would receive, and how much of their findings could be made public.