Third-party evaluators report flaws and potential vulnerabilities in OpenAI’s AI monitoring system after reviewing it as part of a pilot test. Channel NewsAsia says the external review finds issues in how the monitoring system functions, indicating gaps that need attention.
CNBC adds that OpenAI’s response includes plans to improve oversight by involving independent evaluators more directly. It reports that OpenAI CEO Sam Altman pledges to give independent evaluators “employee-like access” to the company, aiming to help address safety concerns that have been increasingly raised. Taken together, the outlets focus on both the evaluators’ findings during the pilot and the steps OpenAI proposes to enable more thorough third-party assessment, reflecting a tension between technical scrutiny and access needed to verify fixes.