The rift between the US Pentagon and Anthropic is not yet final. In a last-ditch effort, Anthropic is attempting to stay on good terms with the Pentagon.
Last Friday, the US Department of Defense (or Department of War since Trump took charge) ruthlessly sidelined Anthropic following an ethical dispute. Anthropic not only lost its exclusivity within the US government but was even labeled a national security risk. OpenAI saw its chance and took over the contract in no time.
Despite the many harsh words exchanged between both camps, the split is not yet final. Anthropic CEO Dario Amodei is reportedly back at the negotiating table at the Pentagon, according to the Financial Times. Amodei hopes to still find a compromise regarding the military use of his AI models.
Ethical discussion
This is precisely where Anthropic clashed with the Pentagon. Anthropic was the department’s exclusive AI partner but wanted to impose certain conditions in the contract that the Pentagon did not agree to. From an ethical standpoint, the conditions sounded reasonable: no mass surveillance or development of autonomous weapons.
However, Pentagon chief Pete Hegseth did not believe Anthropic should be lecturing him. The government would determine for itself what is and isn’t permissible with AI technology. US law also imposes no strict limitations on the use of AI in a military context. Anthropic was given until last Friday to concede, but that did not happen.
The consequences were severe for Anthropic. The government promptly blacklisted Anthropic and labeled the company a ‘security risk’—a dishonor usually reserved for companies from hostile states. Ironically, its replacement, OpenAI, managed to secure the very guarantees that Anthropic had requested.
Mutual accusations
Now Anthropic is trying to get a foot back in the door with the government, but whether the negotiations will proceed in an friendly atmosphere is doubtful. Heavy mutual accusations have been heard over the past week. Emil Michael, the Pentagon under-secretary with whom Amodei will have to negotiate, recently called the Anthropic CEO a ‘liar with a God complex.’
But Amodei has not held back either. In an internal memo to employees, which was made public by The Information, Amodei wrote that he found the whole affair suspicious and that the communication from the Pentagon and OpenAI is full of lies. Anthropic was allegedly sidelined because “we haven’t given Trump dictatorial-style praise,” according to Amodei.
Anthropic declined to comment further on the Financial Times article. There are literally millions at stake for the company: the AI contract with the Pentagon was valued at 200 million dollars.