OpenAI has floated the idea of creating a global AI governance body led by the US and including China as a member, according to Reuters reports. OpenAI's Vice President of Global Affairs Chris Lehane said Wednesday that the US has an opportunity to use its lead in AI technology to create a global governance mechanism resulting in safer, more resilient systems. Such an organization could resemble the International Atomic Energy Agency, which includes China and sets global safety standards for nuclear energy development to prevent weapons proliferation. Lehane suggested connecting the US Commerce Department's Center for AI Standards and Innovation with AI safety institutes that are being created around the world as one way to establish this framework. However, it's unclear whether the Trump administration would welcome global guidelines that include China, as White House officials have previously indicated they would reject the idea of worldwide governance over AI technology. OpenAI has also called for the US to require government researchers to evaluate cutting-edge models, so their safety can be tested in a classified setting, adding another layer to the proposed governance framework.
US and Chinese delegations have convened in Beijing to outline safety protocols for artificial intelligence, marking a significant advancement in the ongoing discussions between the world's two AI superpowers. As reported by Reuters, the summit underscores a shared intent to ensure AI's secure deployment while safeguarding innovation. The meetings follow revelations of software vulnerabilities in Anthropic's Mythos AI tool, prompting urgent security updates from banks and other firms. Government officials warn of potential disruptions to financial markets by criminal or terrorist elements, highlighting the urgent need for coordinated action on AI security. Nvidia CEO Jensen Huang, whose company's semiconductor chips power much of the AI ecosystem, is in Beijing for the summit, reflecting the high stakes involved in AI security coordination. Trump landed in Beijing earlier on Wednesday local time for the first state visit to China by a US leader in nine years, with his entourage including more than a dozen US business leaders such as Nvidia Corp chief executive officer Jensen Huang, who was a last-minute addition to the trip. Before the US delegation departed, officials said they would explore the possibility of opening a new channel of communication with China to regularly discuss AI issues, indicating a more formalized approach to ongoing cooperation.
US Treasury Secretary Scott Bessent has confirmed that Washington is actively engaging with China on establishing AI guardrails and best practices to prevent advanced artificial intelligence models from falling into non-state actor hands. Speaking in a pre-recorded interview, Bessent emphasized that the US maintains an advantage in the AI race while acknowledging China's growing capabilities. As reported by CNBC, Bessent stated that "we are going to discuss AI guardrails with China" and expressed comfort with these discussions because the US remains ahead in the AI race. The Treasury Secretary noted that China is substantially behind the US on AI but has advanced capabilities in the field. The U.S. has cleared around 10 Chinese companies to buy Nvidia's second most-powerful H200 chip, with Bessent telling CNBC that such decisions would be made by the U.S. Commerce Department, not the Treasury. The U.S. government considers the 'big three AI players' - OpenAI, Google and Anthropic - to be strong partners in shaping responsible AI development. One possible point of discussion is a broader effort by the US to address complaints from US AI companies, including OpenAI and Anthropic, that Chinese developers are unfairly using the outputs of cutting-edge American models to produce rival systems at a fraction of the cost and with fewer guardrails.
The summit includes industry leaders like Nvidia's CEO Jensen Huang and xAI's Elon Musk attending the discussions, reflecting the high stakes involved in AI security coordination. Bessent highlighted the rapidly growing capabilities of AI systems being developed by major companies including OpenAI, Google and Anthropic. The Treasury Secretary stressed that while safeguards are necessary, the Trump administration does not want to 'stifle AI innovation' during these discussions. According to Reuters, Bessent explained that "our responsibility is to come up with the highest performance calculus where we can get the most innovation and the highest level of safety." The US approach focuses on promoting US-led standards for global AI governance, with Bessent stating that "we're going to put in U.S. best practices, U.S. values on this, and then roll those out to the world." The Trump administration has been preparing an executive order on AI cybersecurity that emphasises voluntary rather than compulsory pre-deployment model review, indicating a more collaborative approach to AI safety standards.
The discussions come amid growing concern in Washington over increasingly powerful AI models, with Anthropic's new 'Mythos' model having raised alarms due to its alleged advanced cyberattack capabilities. According to Reuters, the talks are focused on developing protocols and best practices to limit the risks associated with advanced AI systems while preserving innovation. The summit's focus on non-state actor prevention represents a direct response to concerns about criminal or terrorist groups exploiting frontier AI systems. The U.S. Treasury has been working with the 11 largest banks to ensure that they patch vulnerabilities and will do the same with smaller 'super-regional' banks and with community banks, with Bessent expressing confidence in a 'smooth transition' to AI's new capabilities. Bessent hinted at major advancements in upcoming large language models from Google Gemini and OpenAI, suggesting continued innovation despite security concerns and the need for coordinated global responses to emerging AI threats.
