February 24th, according to Axios, oneUSADefense officials confirm that Elon Mask's artificial intelligence company xA An agreement has been signed to allow the United States military to use its large model in the confidential system Grok.
To date, only the Anthropic Claude model is available in the United States military's secret system of handling top-secret intelligence, weapons development and combat operations. But the Pentagon is putting pressure on Anthropic on security restrictions and may soon need to find alternatives。
Anthropic rejected the Pentagon’s request to open Claude to “all legitimate uses” and insisted, inter alia, on banning the use of models for mass surveillance of American citizens and for the development of fully autonomous weapons。
As previously reported by Axios, XAI has agreed to this “all legitimate uses” criterion. The New York Times previously reported that the parties had signed an agreement. XAI did not respond to the request for comment。
It is not clear whether xAI will be a complete replacement for Anthropic and how long this process will take。
According to information from IT House, United States Secretary of Defense Peter Hegesses will meet with the CEO of Anthropic Dario Amordi at the Pentagon on Tuesday, local time, and sources expect the meeting to be tense。
An official of the Ministry of Defence indicated that Hegesses would in fact give Amody an ultimatum. If Anthropic does not agree to lift all restrictions, the Pentagon will threaten to classify them as “supply chain risk” and may impose other penalties。
U.S. Department of Defense officials recognize that moving and replacing Claude would be an extremely difficult task。
Grok, Google Gemini, ChatGPT from OpenAI can all be used in military non-confidential systems; Google and OpenAI have also started negotiations on entering the field。
The Pentagon is accelerating these negotiations in preparation for a possible break-up with Anthropic。
According to information, the Pentagon has “engaged with a new sense of urgency to restart the negotiations” but, due to the complexity of the issue, the parties “are far from reaching agreement”. Another source confirmed that contacts in the Pentagon were accelerating。
According to The New York Times, Google “close” to an agreement to allow Gemini to be used in a secret setting, while OpenAI “has no eyeballs”。
The Department of Defense official challenged the claim that Google was much faster than OpenAI, stating that negotiations were ongoing with the two companies and that the United States Department of Defense was confident that the parties would eventually sign the agreement。
However, United States government officials insist that Google and OpenAI agree to the criterion of “all legitimate uses”. According to sources, it is not clear whether OpenAI will accept this standard, but “negotiations are still under way”。