The end of the Anthropic monopoly in the US military? OpenAI and Google are waiting to enter classified systems

The Pentagon is close to terminating cooperation with Anthropic, the creator of the AI ​​Claude model, and putting it on a “blacklist” because of its desire to limit the military use of the technology, Axios reported on Monday. Claude was to be used during the operation to capture Nicolas Maduro.

The portal writes, citing a Pentagon official, that the Ministry of Defense is threatening to give Anthropic the status of a “supply chain risk”, which would force all Pentagon contractors to dissociate themselves from this company's products. As Axios notes, such status is usually reserved for foreign adversaries of the US, such as Chinese technology companies.

Pentagon spokesman Sean Parnell confirmed that The ministry's relationship with Anthropic is “under review”.

Our country requires partners to be ready to help our soldiers win every fight, he said.

The role of AI in the operation against Maduro

The source of the dispute are ongoing negotiations regarding the terms of use of the Claude model by the army. According to the website, Claude is currently the only artificial intelligence model available in the US military's classified IT systems and was reportedly used during the operation against Venezuelan President Nicolás Maduro in January this year.

An Axios insider admitted that breaking up with the company will be very complicatedand the company offers the best services in the industry, but he noted that the ministry wants it to “pay the price for enforcing” restrictions.

Where is the limit of surveillance?

The company's president, Dario Amodei, expressed his readiness to ease the current conditions of use, but demands a guarantee that his tools will not be used for mass surveillance of Americans or to create fully autonomous weapons systems operating without human intervention.

The Pentagon finds these restrictions unacceptable and insists that AI can be used for “all lawful purposes.”

An Anthropic spokesman told Axios that the company is in “good faith, productive discussions” with the ministry. Another company representative pointed out that current law on mass surveillance does not take into account the capabilities of AI.

In a recent interview with the New York Times, Amodei noted that using AI for surveillance could radically violate people's privacy and called for a law or even a constitutional amendment to address these risks.

Negotiations with the giants of Silicon Valley

According to Axios, the dispute with Anthropic sets the tone for the Pentagon's negotiations with three other large AI labs — OpenAI, Google and xAI — that have agreed to remove safeguards on classified military systems but do not yet operate on classified systems.

From Washington Oskar Górzyński (PAP)

osk/mal/