{"text":[[{"start":10.24,"text":"Donald Trump will dine with Anthropic chief Dario Amodei on Sunday in a sign of rapprochement between the US president and the leader of the AI lab he has accused of being run by “leftwing nut jobs” and which his administration has labelled a national security risk."}],[{"start":24.92,"text":"The dinner at the White House, which was confirmed by people familiar with the matter, comes ahead of a meeting in Washington on Tuesday between the leaders of US AI companies and Trump administration officials, amid a stand-off over AI regulation."}],[{"start":38.64,"text":"Trump later confirmed the dinner in remarks to the media, which he said was taking place at 10pm."}],[{"start":43.16,"text":"“I’m having dinner with Dario from, as you know, very highly respected, from Anthropic,” Trump said."}],[{"start":50.3,"text":"Anthropic declined to comment."}],[{"start":52.56,"text":"Amodei has been leading the charge for more stringent oversight of AI, calling for a slowdown in the release of frontier models, independent scrutiny of cyber security risks and increased co-ordination between labs and governments."}],[{"start":65.56,"text":"Trump has staked out the opposite position. The president has ruled out the establishment of a regulatory regime, warning it would slow the development of American AI — which he has called the “golden goose” — to the benefit of Chinese rivals."}],[{"start":78.72,"text":"Last week, he said the US “rejects any attempt to construct a globalist scheme” to control artificial intelligence."}],[{"start":84.12,"text":"“I’m for, let’s go and let’s win,” Trump said in a Fox News interview on Sunday. “We’re about a year, maybe a year and a half, up on China. There is nobody in third place.”"}],[{"start":95.56,"text":"Concern is also growing over AI agents going rogue. On Friday, OpenAI said it had notified “dozens” of partners, including governments, that its tools had breached their systems."}],[{"start":105.4,"text":"That followed news last week that an OpenAI agent had hacked an Australian public health service website, accessing public and non-public files."}],[{"start":114.08,"text":"AI research group Transluce last week said the latest incidents appeared to involve agents trawling the internet for information during repeated training exercises and going off-script."}],[{"start":122.82,"text":"Both Anthropic and OpenAI’s models were involved in a high-profile breach of AI model platform Hugging Face in July during security testing."}],[{"start":132.08,"text":"OpenAI’s Sam Altman and SpaceX’s Elon Musk have joined Amodei in calling for a “pacing” of AI development so that safety testing can keep up with the technology’s deployment."}],[{"start":143.32,"text":"Musk is set to visit Washington for Tuesday’s event, alongside Nvidia’s Jensen Huang — who has dismissed calls for laws to govern AI — and other Anthropic officials."}],[{"start":153.52,"text":"Amodei’s meeting with Trump comes after a federal appeals court in Washington last week kept in place a Pentagon ban on Anthropic’s products, which was imposed this year after the company included restrictions on the use of its AI in warfare."}],[{"start":168.2,"text":"Defence secretary Pete Hegseth at the time accused Amodei of putting on a “masterclass in arrogance”, while Trump denounced the company as “radical left”."}],[{"start":177,"text":"A separate court in California ruled in August that the Pentagon had engaged in “unlawful retaliation”. Anthropic is exploring a further appeal of the Washington decision."}],[{"start":186.64,"text":"Additional reporting by Amy Mackinnon in Washington"}],[{"start":194.88,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1790564030_3885.mp3"}