Regardless of Grok's 'MechaHitler' Meltdown, Elon Musk’s xAI Scores $200M Pentagon Deal – Decrypt

Not even a “MechaHitler” meltdown was sufficient to scare off the Pentagon, apparently.

Simply days after xAI’s Grok chatbot generated a stream of antisemitic and homophobic responses, Elon Musk’s firm introduced that it secured a $200 million protection contract to develop and provide AI instruments to federal staff.

The Division of Protection’s Chief Digital and Synthetic Intelligence Workplace introduced the contract on Monday, saying it’s designed to assist the federal government faucet into the most recent AI instruments for every part from battlefield operations to enterprise wants.

“The adoption of AI is remodeling the Division’s potential to help our warfighters and preserve strategic benefit over our adversaries,” CDAO Chief Digital and AI Officer Dr. Doug Matty stated in an announcement. “Leveraging commercially accessible options into an built-in capabilities method will speed up the usage of superior AI as a part of our joint mission important duties in our warfighting area in addition to intelligence, enterprise, and enterprise data techniques.”

In accordance with Lucas Hansen, the co-founder of the AI safety and training nonprofit CivAI, the cope with xAI is much less about fast use and extra about preserving tempo with world rivals.

“The federal government needs to help the frontier labs and growth of AI in order that the U.S. doesn’t fall behind China,” Hansen informed Decrypt. “Then there’s the query of how these funds are distributed and the way that help is offered. I feel this would possibly simply be the precise method they’ve chosen to help the businesses, somewhat than an expectation of inside deployment instantly.”

However Hansen warned that Grok’s underlying conduct raises deeper considerations about the way it would possibly act in high-stakes authorities settings.

“Even when Grok is deployed by the federal government utilizing the identical model that beforehand confirmed a willingness to generate dangerous responses, then—even when these use instances aren’t explicitly antisemitic or hateful—it’s essentially extra prone to behave unethically and autonomously,” he stated.

Grok has been the middle of a number of public meltdowns. In Could, studies surfaced that Grok inserted “white genocide” claims into unrelated prompts. xAI blamed the problem on a rogue worker. Final week, simply earlier than the Grok 4 launch, the chatbot once more drew criticism for a wave of racist and homophobic outputs, resulting in the resignation of X CEO Linda Yaccarino.

Whereas considerations linger over how Grok would possibly behave in delicate environments, Hansen stated the model the Pentagon makes use of will not be prone to be the identical because the one which X customers have had entry to since November 2023.

The $200 million award is the most recent deal between the USA navy and prime AI builders. Different AI builders receiving $200 million contracts embody Google, Anthropic, and OpenAI. In November, Meta stated it could make its open-source Llama AI mannequin accessible to U.S. protection companies and contractors. In June, ChatGPT developer OpenAI was awarded its personal $200 million contract with the Division of Protection, launching its AI suite, OpenAI for Authorities.

If the Pentagon deployed Grok over a rival, Hansen emphasised it could be as a result of Grok is “open-weight,” that means it may be run on non-public {hardware} and never require an API connection again to xAI.

Regardless of Grok’s latest controversies, on the coronary heart of the deal, Hansen stated, could also be a easy strategic calculation: Keep near the important thing gamers shaping the way forward for AI.

“I feel most of that is the federal government eager to have a relationship with all of the frontier labs, and so they hit the 4 labs that really matter,” he stated.

In a separate assertion asserting the cope with the Pentagon, xAI stated its mission is to create AI instruments to “help humanity in our quest for understanding and information.” To do that, xAI introduced the launch of Grok for Authorities, a collection of AI instruments together with the lately launched Grok 4 and Deep Search.

“America is the world chief in AI, and that is in no small half as a result of a convention of innovation and robust investments in engineering and science,” xAI wrote. “We’re excited to contribute again to the nation that made xAI uniquely attainable right here.”

xAI didn’t reply to requests for feedback from Decrypt.