Anthropic began last-ditch talks to save its deal with the US Department of Defense and avoid being designated a supply-chain risk.
The firm’s relationship with the Pentagon broke down over the use of its AI products in domestic surveillance and fully autonomous weapons, but its chatbot Claude is still being used to plan attacks on Iran.
The negotiations may be hindered by an internal memo from the Anthropic CEO saying it had been frozen out for not giving “dictator-style praise to [US President Donald] Trump.” It has also been hampered by its own investors failing to back it, Semafor’s technology editor reported. Meanwhile, talks began in Geneva between 128 nations to establish an international framework regulating the use of semi-autonomous weapons.