Pentagon starts testing new AI models to replace Anthropic Claude
The US Pentagon has reportedly started testing new AI tools from the likes of OpenAI and Google as it stops using Anthropic's models. This comes after the US Department of Defense had a fallout with the Dario Amodei-led firm over unrestricted AI use. Here is everything you need to know.
A few weeks ago, US Defense Secretary Pete Hegseth designated Anthropic a “supply chain risk” after a fallout over unrestricted AI use. Anthropic decided to walk away from a multi-million dollar deal for the use of its Claude models by the Pentagon. With the latter signing a new deal with OpenAI instead. Now, a new report suggests that the Pentagon is testing new AI models that can serve as the alternative to Anthropic’s AI tools.
As per a report from Bloomberg, the US Department of Defense began testing new AI models from the likes of Google and OpenAI in March this year – three days after Hegseth labelled Anthropic a supply chain risk. The Pentagon is reportedly working with 25 of the department's “power users” as it looks for alternatives to Claude.
Keep in mind that the Pentagon had signed a deal with OpenAI just hours after its fallout with Anthropic. A deal which OpenAI CEO Sam Altman later admitted was “rushed.”
Anthropic’s Claude is heavily used on the military's Maven Smart System for classified operations against Iran. Previous reports claimed that the use of AI could’ve been a reason behind US’ strikes on an Iranian school in February that killed 165 schoolgirls.
Why is Pentagon looking for alternatives to Claude?
The fallout between the Pentagon and Anthropic happened largely due to two conditions imposed by Anthropic. The Pentagon wanted access to Claude for “all lawful” use cases. However, Anthropic had two conditions – the AI cannot be used for mass domestic surveillance in the US, and it cannot be used for the development of autonomous weapons.
Ultimately, Anthropic backed away from the deal entirely and the Pentagon gave itself six months to wind down use of the company's products. Since then, Anthropic has challenged the supply chain designation – a label that is usually used for foreign adversaries like Huawei – in court.
While there have been signs of de-escalation between the White House and Anthropic after the release of Mythos, a new model described as a threat to existing cyber-security defences, reports indicate that Pentagon officials have shown no sign of repairing ties with the company.
US Defense Undersecretary Emil Michael told Bloomberg that he expected rival developers to produce models with similar capabilities to Anthropic's most advanced systems every month or two.
US military wants new AI models
As per reports, the Pentagon is testing models from companies like Google and OpenAI in a digital platform separate from the Maven Smart System. The new evaluation is being carried out by so-called power users across five military theatre commands worldwide.
Although some users were expected to favour Anthropic, the Pentagon has not seen the level of pushback it had anticipated over trying other models. Officials want operators to have access to several models so that no single company can find itself in a position of leverage.
A few weeks ago, US Defense Secretary Pete Hegseth designated Anthropic a “supply chain risk” after a fallout over unrestricted AI use. Anthropic decided to walk away from a multi-million dollar deal for the use of its Claude models by the Pentagon. With the latter signing a new deal with OpenAI instead. Now, a new report suggests that the Pentagon is testing new AI models that can serve as the alternative to Anthropic’s AI tools.
As per a report from Bloomberg, the US Department of Defense began testing new AI models from the likes of Google and OpenAI in March this year – three days after Hegseth labelled Anthropic a supply chain risk. The Pentagon is reportedly working with 25 of the department's “power users” as it looks for alternatives to Claude.
Keep in mind that the Pentagon had signed a deal with OpenAI just hours after its fallout with Anthropic. A deal which OpenAI CEO Sam Altman later admitted was “rushed.”
Anthropic’s Claude is heavily used on the military's Maven Smart System for classified operations against Iran. Previous reports claimed that the use of AI could’ve been a reason behind US’ strikes on an Iranian school in February that killed 165 schoolgirls.
Why is Pentagon looking for alternatives to Claude?
The fallout between the Pentagon and Anthropic happened largely due to two conditions imposed by Anthropic. The Pentagon wanted access to Claude for “all lawful” use cases. However, Anthropic had two conditions – the AI cannot be used for mass domestic surveillance in the US, and it cannot be used for the development of autonomous weapons.
Ultimately, Anthropic backed away from the deal entirely and the Pentagon gave itself six months to wind down use of the company's products. Since then, Anthropic has challenged the supply chain designation – a label that is usually used for foreign adversaries like Huawei – in court.
While there have been signs of de-escalation between the White House and Anthropic after the release of Mythos, a new model described as a threat to existing cyber-security defences, reports indicate that Pentagon officials have shown no sign of repairing ties with the company.
US Defense Undersecretary Emil Michael told Bloomberg that he expected rival developers to produce models with similar capabilities to Anthropic's most advanced systems every month or two.
US military wants new AI models
As per reports, the Pentagon is testing models from companies like Google and OpenAI in a digital platform separate from the Maven Smart System. The new evaluation is being carried out by so-called power users across five military theatre commands worldwide.
Although some users were expected to favour Anthropic, the Pentagon has not seen the level of pushback it had anticipated over trying other models. Officials want operators to have access to several models so that no single company can find itself in a position of leverage.
