AI, Trump asks companies to share advanced models in advance
The new order stems above all from the concerns raised by Mythos, the new generation model developed by Anthropic
-U70008020720VNb-1440x752@IlSole24Ore-Web.JPG?r=650x341)
Donald Trump has signed a new executive order on artificial intelligence that introduces a system of voluntary controls on the most advanced models, while scaling back the measures envisaged in the draft he had blocked a few weeks ago. The measure,signed confidentially without the initially planned public ceremony, as Politico explains, represents the latest chapter in the confrontation that has been dividing the White House for months on how to deal with the risks and opportunities of artificial intelligence.
According to Politico, Trump signed the order after a closed meeting held at the White House the previous day, which was attended by some of the key administration officials involved in the dossier. The decision comes after a sudden turnaround on 21 May, when the president had cancelled at the last moment the signing of a stricter version of the measure, which had already been approved at the highest levels of the White House and viewed by companies such as OpenAI, Google and Anthropic.
At the centre of the dispute was the length of the prior review period for new artificial intelligence models. The original draft required companies to submit their systems to the government for up to 90 days before public release.The final version reduces the deadline to a maximum of 30 days and maintains the voluntary nature of the procedure. Companies will be asked to share their models in advance so that the authorities can assess their impact on cybersecurity, critical infrastructure, the financial system and national security.
In the text of the order Trump states that advanced artificial intelligence capabilities strengthen the United States but also introduce new national security considerations. The president also reiterates the goal of maintaining US leadership in the field, explicitly linking the new policy to strategic competition with China.
Behind the compromise lies the internal battle that runs through the administration. On one side are supporters of more controls, worried by the technological acceleration of recent months. On the other are exponents who fear over-regulation. These includeDavid Sacks, former head of White House policy on artificial intelligence, who according to Politico warned Trump shortly before the planned May signing that the measure would slow down American innovation. Sacks also attended the preparatory meeting that preceded the approval of the final text.
