{"text":[[{"start":9.151,"text":"The US government has lifted its export controls on Anthropic’s leading models, ending a weeks-long impasse between the $1tn AI company and the Trump administration."}],[{"start":19.27,"text":"The US Department of Commerce told Anthropic it had lifted the ban on foreigners accessing the company’s Mythos and Fable models on Tuesday evening, according to people with knowledge of the matter. The decision allows the AI group to re-release its latest model, Fable 5, to the general public."}],[{"start":35.54,"text":"The government allowed Anthropic to re-release Mythos 5, a model with fewer safeguards aimed at enterprise customers, to 100 or so pre-vetted partners last week."}],[{"start":45.16,"text":"In a letter to Anthropic co-founder Tom Brown seen by the FT, commerce secretary Howard Lutnick said he was removing the ban after Anthropic had “agreed to proactively detect and address security risks associated with the models”."}],[{"start":58.14,"text":"A person close to Anthropic said the company had “implemented a new safeguard” which specifically addresses the concerns that led to the ban. That safeguard has been tested and approved by the government’s Center for AI Standards and Innovation, they added."}],[{"start":72.5,"text":"Lutnick had placed export controls on Anthropic on June 12, after the government discovered a “jailbreak”, or a way of bypassing safety guardrails built into the model. That led Anthropic to pull its powerful models for all users worldwide."}],[{"start":86.48,"text":"The episode inflamed tensions between the White House and Silicon Valley, which has criticised the government’s ad hoc approach to regulating frontier AI. It also led to a backlash from foreign governments, which accused the US of punishing its allies by imposing the ban."}],[{"start":101.9,"text":"The export control came after the Trump administration — under pressure from the AI lobby — unveiled a voluntary framework for monitoring cutting-edge models, and vowed not to impose a regulatory regime on leading AI companies."}],[{"start":115.54,"text":"But several senior figures in the US government have warned of the risks of releasing such models to the public before their potential to exploit cyber security vulnerabilities in critical industries is properly examined."}],[{"start":128.02,"text":"Anthropic, whose models have propelled the start-up ahead of rivals such as OpenAI, was the first AI group to be hit by a White House ban after it triggered concerns that its technology could be used to exploit cyber security vulnerabilities."}],[{"start":140.66,"text":"OpenAI has also faced restrictions on its new model, GPT-5.6. The San Francisco-based group was asked by the Trump administration to initially limit access to around two dozen government-approved partners."}],[{"start":155.76,"text":"A wider release of GPT-5.6 could come as early as next week, according to a person with knowledge of the matter."}],[{"start":163.18,"text":"Meanwhile, the US and Europe have discussed creating a “trusted partner” scheme for cutting-edge AI models, under which close US allies would receive privileged access to the latest tools."}],[{"start":174.82,"text":"Anthropic declined to comment."}],[{"start":177.14,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1782889139_9451.mp3"}