Global efforts to establish trustworthy AI regulation

Industry leaders, policymakers, and businesses are racing to develop trustworthy AI regulation amid an ad hoc regulatory framework and delayed federal action in the U.S. AI CEOs are navigating the challenge of testing advanced systems and aligning with Beijing on safety standards, while the White House advocates for self-regulation by the industry. A White House official emphasized that companies possess the most expertise in AI and understand the requirements of the voluntary framework. However, friction arises as third-party evaluators, such as those at METR, face scrutiny for their ties to the effective altruism movement and the companies they oversee. OpenAI recently released a public incident reporting playbook following six new incidents. SaferAI's Henry Papadatos stressed that public transparency is crucial, but voluntary efforts lack sufficient enforcement. Treasury Secretary Scott Bessent hinted at potential safety discussions with Chinese Vice Premier He Lifeng during upcoming talks. Elon Musk proposed cross-review between U.S. and Chinese labs, though many view this as improbable due to intense competition. Existing AI evaluation systems have roots in marketing new models, including performance on coding and math benchmarks. Despite progress, experts caution that voluntary measures alone are insufficient to ensure AI safety and trustworthiness.