{"text":[[{"start":6.9,"text":"The US government is in advanced talks with AI companies to create voluntary standards for the release of new models after intervening in the rollout of state of the art tools from Anthropic and OpenAI."}],[{"start":18.15,"text":"The standards, to be announced as soon as next week, would set benchmarks for models with cutting-edge cyber capabilities and establish release timelines in an effort to streamline future launches, according to people familiar with the talks. "}],[{"start":31.599999999999998,"text":"The move would put into effect plans laid out in President Donald Trump’s executive order on AI last month, as groups seek clarity on how governments will regulate the technology ahead of expected blockbuster public listings from OpenAI and Anthropic."}],[{"start":48.25,"text":"US officials and companies accelerated their talks after a series of interventions from the Trump administration last month, which sparked confusion and anger within leading labs, the people added."}],[{"start":60.95,"text":"Anthropic was hit with export controls on its latest models on June 12 over concerns about its susceptibility to misuse and cyber security risks. The restrictions were lifted on Tuesday. "}],[{"start":72.65,"text":"One person familiar with the White House discussions said the Center for AI Standards and Innovation and the National Security Agency would play a crucial role in setting and monitoring the standards."}],[{"start":84,"text":"US officials would also clarify who is able to access advanced models, both domestically and abroad, in a move that could set the stage for a global framework including US allies, said one of the people. "}],[{"start":95.75,"text":"Negotiations between industry and the White House have included the duration of model reviews, as well as the threshold at which models are considered “frontier” — or having the greatest capability and by implication the greatest risk. "}],[{"start":109.3,"text":"Technical teams from the labs had been meeting with government officials on a regular basis over the past week, added the person. That has put the White House on course to release its standards as early as next week, though that schedule might slip if officials struggle to reach a consensus, said multiple people with knowledge of the talks."}],[{"start":128.7,"text":"The White House did not immediately respond to a request for comment."}],[{"start":132.85,"text":"AI groups including OpenAI and Anthropic made voluntary commitments under President Joe Biden in 2023 to test the safety of their models ahead of public release. Since Trump’s re-election in 2024, the president has promoted a light-touch regulatory regime in an effort to enable US companies to innovate faster than their Chinese rivals. "}],[{"start":155.45,"text":"But the release of increasingly sophisticated AI models, capable of exploiting cyber security vulnerabilities and potentially disrupting critical industries, has tested the limits of that approach. "}],[{"start":167.39999999999998,"text":"In his June 2 executive order, Trump laid out plans “to develop and maintain a classified benchmarking process to assess the advanced cyber capabilities of AI models”. "}],[{"start":177.39999999999998,"text":"After the restrictions on Anthropic’s latest models were lifted on Tuesday the company said it would “work with the government and with industry peers [including Amazon, Microsoft and Google] towards a shared, voluntary security and evaluation standard for frontier model providers”."}],[{"start":193.2,"text":"The Trump administration has also asked OpenAI to release its GPT-5.6 only to groups vetted by the administration, with a wider release expected as soon as next week."}],[{"start":204.25,"text":"“We appreciate the administration’s ongoing engagement as we work toward a broad release of GPT-5.6 and a durable framework for future frontier model releases under the recent executive order,” an OpenAI spokesperson said on Wednesday."}],[{"start":219.1,"text":"Google has been in discussions with the government ahead of releasing its own advanced coding models, which will have more sophisticated cyber capabilities than prior generations, according to two people with knowledge of the matter. The tech giant was also involved in wider discussions about industry standards, they added."}],[{"start":237.45,"text":"OpenAI, Anthropic and Google declined to comment specifically on talks regarding the new standards."}],[{"start":243.64999999999998,"text":"Senior figures in the industry as well as in Washington have expressed fear that the White House’s piecemeal intervention so far risks slowing American innovation and ceding the advantage to China. "}],[{"start":255.59999999999997,"text":"Sam Altman, OpenAI’s chief executive, called for a “global framework” to establish “standards [and] provide expert and impartial analysis of capabilities and risks”, in the FT. Only after such a framework was in place could the benefits of AI be broadly distributed, he said."}],[{"start":273.99999999999994,"text":"Additional reporting by Cristina Criddle"}],[{"start":282.94999999999993,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1782959727_3926.mp3"}