{"text":[[{"start":10.95,"text":"The Trump administration has asked OpenAI to stagger the release of its state of the art AI model so that organisations set to gain early access to the powerful tool can first be vetted by the US government."}],[{"start":23.799999999999997,"text":"OpenAI plans to give about two dozen partners access to its latest model, GPT 5.6, ahead of a broader rollout, according to multiple people with knowledge of the matter. Each of those partners would have to be approved by the government, they added."}],[{"start":37.599999999999994,"text":"The company is taking the staggered approach amid fears about the ability of powerful new models to breach existing cyber security defences, disrupt critical industries and leak to adversarial countries, including China."}],[{"start":51.14999999999999,"text":"Requests to limit distribution came from the US Treasury department, the commerce department and White House’s Office of the National Cyber Director and Office of Science and Technology Policy, they added."}],[{"start":63.19999999999999,"text":"The intervention is the latest in the administration’s piecemeal effort to harness the power of the technology while limiting its risks. The White House has set out plans to develop a voluntary framework for ensuring new models are safe before release but stopped short of implementing an official process."}],[{"start":80.99999999999999,"text":"President Donald Trump has also warned against heavily regulating AI and his advisers have insisted companies “must be free to innovate”."}],[{"start":90.29999999999998,"text":"For now, the government has weighed in on an ad hoc basis to contain the risks from new AI models that exhibit powerful capabilities. "}],[{"start":98.79999999999998,"text":"Earlier this month, the Department of Commerce imposed export controls on OpenAI’s rival Anthropic, barring foreign nationals from using its newest models, Fable and Mythos, due to security concerns. That prompted Anthropic to suspend the models for all users. "}],[{"start":115.49999999999999,"text":"The US government had been examining Mythos for weeks before the release, with Treasury secretary Scott Bessent raising concerns about the model’s ability to exploit vulnerabilities in the global financial system."}],[{"start":127.09999999999998,"text":"Several other companies from a cross-section of US industry were given early access to Mythos so they could identify and patch any security risks posed by the tool prior to its wider release."}],[{"start":138.59999999999997,"text":"The commerce department’s move to impose export controls was prompted by a report from researchers at Amazon, who identified a potential “jailbreak” — a way of bypassing the model’s guardrails and extracting information about security vulnerabilities in software."}],[{"start":153.49999999999997,"text":"Amazon, a big investor in Anthropic, first shared its findings with the AI company before its chief executive Andy Jassy raised broader concerns about frontier-model capabilities with US officials."}],[{"start":166.19999999999996,"text":"Negotiations between Anthropic and the US over reinstating access to Mythos are ongoing, according to people familiar with the matter."}],[{"start":175.04999999999995,"text":"Anthropic and the government clashed earlier in the year over the company’s refusal to allow unrestricted use of its technology in warfare. A federal judge has blocked the Pentagon from placing sanctions on the company over the move but a legal fight between the pair is ongoing. "}],[{"start":192.44999999999996,"text":"The Information first reported that OpenAI was staggering the release of its newest model. The White House did not immediately respond to a request for comment. OpenAI declined to comment. "}],[{"start":212.94999999999996,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1782443696_5457.mp3"}