w////am, 🐝

Quick take from me on what US midterms mean for the AI regulation debate here in the UK: Critically, Trump has said he accepts the election results, and that he wants to work with Democrats. AI regulation is going to be high on the congressional Democrats’ agenda, and there is a decent chance Trump will work with them on a package rather than seek to start vetoing legislation so early. Congressional Republicans, seeing how unpopular data centres and Big Tech generally were in this election, may even come to the table as well. I sincerely hope that this frees up space for UK ministers to start more purposefully thinking about the ways in which we need to be changing the law to protect our society from some of the ramifications this technology. To date, UK ministers have opposed an AI Bill every step of the way, seeking instead to maintain @AISecurityInst’s access to frontier models (and the special relationship more broadly, wanting to avoid a divergence at all costs). They are reported to have sought to avoid a ‘Christmas-tree’ bill that sees MPs submitting hundreds of amendments seeking to constantly expand the scope of the bill. But the way to avoid a Christmas-tree AI Bill is to bring forward a government bill that meets the scale of the moment, covering areas as wide as frontier safety, liability, copyright, employment rights, public sector decision-making, online safety, data protection and elections. The case for continued total deference to the US is significantly weakened. We must use our hosting of the G20 as an opportunity to take the lead on regulating AI and agree a common framework with international partners. These discussions can take place in consultation with congressional Democrats as they work on their own package and negotiate with Trump.