The Elon Musk’s artificial intelligence (AI) company, Grok, was used in attacks against Iran, the US Administration revealed in a court document. The revelation appears in a document from the United States Department of Justice, which defends the gas turbines of a gigantic data center belonging to the multimillionaire’s company, xAI, the target of an environmental complaint.
In the document, presented on June 15 and consulted by France-Presse (AFP), the Department of Justice argues that the complaint “threatens the national, economic and energy security” of the country, by risking cutting off the power supply to AI infrastructures now used by the army.
To support the argument, the department presents the testimony of Cameron Stanley, responsible for AI at the Pentagon, who declares under oath that a tool derived from Grok, the “Grok Gov Model”, is already being used within the scope of ‘Project Maven’, the army’s AI-assisted targeting program, initially based on Anthropic’s Claude model.
According to the statement, Maven processes “allowed US forces to launch more than 2,000 munitions against 2,000 different targets in 96 hours” during the war against Iran. The senior official sees in these numbers “proof of a very significant increase in operational effectiveness, made possible by the Grok Gov Model”, without specifying whether this is the only model used by the program.
Stanley also states that Maven users consume “around two billion tokens” (computing units) “per day”, which is equivalent to “up to six million pages” processed, a volume that, according to him, makes xAI’s computing infrastructure indispensable.
The turbines targeted by the complaint power Colossus 2, an xAI supercomputer that powers Grok, on the outskirts of Memphis.
The NAACP (National Association for the Advancement of Colored People) sued xAI, accusing it of operating, without a license, dozens of turbines, in violation of the air quality law.
The turbines pollute majority-black neighborhoods, the association says. xAI argues that its turbines are temporary and mobile and, therefore, are not subject to the aforementioned regulations.
At the end of February, the Donald Trump Administration terminated contracts with Anthropic, which refused to allow its tools to be used for fully automated attacks or mass surveillance of Americans.
The Pentagon then turned to Google, OpenAI and xAI, the other three leading American AI companies. The transition, however, takes time and Washington had to admit, in March, that the Claude continued to be used in the war in Iran.
The military use of AI raises controversy. At Google, more than 600 employees demanded in April that AI not be provided to the army for confidential operations. The group had already abandoned the Maven project in 2018, under pressure from its engineers.
Close to Donald Trump, whom he supported and advised, Musk integrated xAI into his SpaceX space group in February, which carried out, on June 12, the biggest entry into the stock market in history.
Source: www.noticiasaominuto.com.br
Source link
