Anthropic CEO Dario Amodei has strongly criticised OpenAI’s agreement with the US Department of Defense over the use of artificial intelligence. In an internal memo, he accused the Sam Altman-led company of making misleading statements and said its claims were untrue. Amodei argued that OpenAI tried to present itself as a stabilising force after the Pentagon’s disagreement with Anthropic.
According to a report by The Information, Amodei described OpenAI’s portrayal of the Pentagon contract as completely false and dismissed its safety commitments as mere “safety theatre.”
OpenAI finalised the deal with the US Department of Defense only hours after Anthropic’s contract was cancelled. Sam Altman later acknowledged that the quick agreement with the Pentagon appeared rushed and somewhat opportunistic.
In the memo, Amodei said the main reason OpenAI accepted the Pentagon contract was because the company was not prioritising the prevention of AI misuse, unlike Anthropic. He stated that OpenAI was more focused on appeasing employees, whereas Anthropic declined the deal because it wanted stronger safeguards against potential abuse of AI systems.
Anthropic had earlier said it stepped away from the agreement due to concerns that its technology might be used for domestic mass surveillance or the creation of autonomous weapons. The Pentagon reportedly rejected these restrictions and wanted the freedom to use AI for any “lawful” purpose.
OpenAI has also stated that it would not permit its AI systems to be used for such applications. Sam Altman recently shared a revised version of the agreement highlighting these boundaries.
Amodei also suggested that the US government’s dissatisfaction with Anthropic might be linked to political factors. In the memo, he claimed the company did not contribute to Donald Trump’s election campaign, while OpenAI President Greg Brockman and his wife reportedly donated $25 million to a Trump-linked super PAC.
Anthropic’s contract was eventually terminated by President Donald Trump, after which the company was labelled a supply chain risk. US Defense Undersecretary Emil Michael later criticised Amodei, calling him dishonest and accusing him of having a “god complex.” Michael also said Anthropic had imposed restrictions, such as preventing the military from using publicly available data sources like LinkedIn.
Despite the contract dispute, the US military is reportedly still using Anthropic’s AI technology in operations related to Iran. The government is expected to undergo a six-month transition period while shifting to OpenAI’s models.
Meanwhile, the Information Technology Industry Council, which includes companies such as Google and Nvidia, sent a letter to the Pentagon expressing concern about the use of supply chain risk designations in procurement disputes. The group urged the Defense Department to resolve disagreements through dialogue or by selecting alternative providers rather than applying such labels. Companies classified as supply chain risks are barred from working with US defense contractors.
Following OpenAI’s deal with the Pentagon, the company faced significant criticism online. Many users reportedly uninstalled ChatGPT and moved to Anthropic’s Claude chatbot, leading to a sharp increase in downloads. Uninstallations of ChatGPT rose by nearly 300 percent, while Claude climbed to the top of the US App Store rankings and even experienced outages due to heavy traffic.
Amodei acknowledged the shift in public sentiment in his memo. He said that although Sam Altman was trying to present himself as a mediator and negotiator, many people viewed OpenAI’s agreement with the Defense Department as questionable, while seeing Anthropic more positively. However, he also expressed concern that some OpenAI employees might accept Altman’s explanation and misunderstand the situation.
