Artificial intelligence developer OpenAI has agreed to integrate six evaluators from Apollo Research into its organization to study whether its advanced models might attempt to act deceptively or exhibit subversive tendencies.

However, the arrangement has drawn scrutiny regarding the autonomy of the outside investigators, as constraints placed on their operational access and public reporting limits have sparked concerns about potential oversight limitations.

Additional information regarding the partnership and the findings of the safety evaluations is anticipated as the story develops.