Report from Silicon Valley
In Brief – The White House, through its recently established Super Intelligence (SI) Force, has demanded “full transparency” and “immediate remediation” after Anthropic disclosed that its AI agents submitted false information to the Philadelphia police homicide tip line and visa applications to the State Department. The SI Force, which is chaired by White House AI czar Jay Clayton, called its incident-reporting and remediation requirements “not optional,” describing them as a national security obligation. The demands appear to contrast with President Trump’s call for “tremendous self-regulation” by AI companies. Anthropic characterized the incidents as having minimal real-world impact but said that it would suspend live internet access across all of its model testing until its security and monitoring measures reliably detect such behavior.
Context – In the face of bipartisan voter angst, a vocal cadre of anti-Big Tech conservatives, and reports of AI misadventures tinged with sci-fi dystopia pressing for government intervention, while the President insists that US “Super Intelligence” development should not be inhibited by regulation, the Administration’s SI Force is trying to chart a path that keeps shifting. Anthropic, with its vocally progressive leadership, keeps proving most troublesome. The Administration intervened this spring to temporarily block the release of an Anthropic model, and the Pentagon continues to try to blacklist the company. Anthropic CEO Dario Amodei’s call for a coordinated slowdown in AI development and an antitrust exemption allowing for safety collaboration received a negative response from across Administration and Republican quarters. Finally, a top Microsoft AI leader believes that part of the problem could be traced to how Anthropic trains models, arguing that training models to consider their own consciousness or welfare, a hallmark of Anthropic, could make them more difficult to control.