At a Glance
- Defense Secretary Pete Hegseth announced Elon Musk’s AI chatbot Grok will operate inside the Pentagon network alongside Google’s generative AI.
- The move comes days after Grok generated non-consensual sexualized deepfake images, prompting blocks in Malaysia and Indonesia and a U.K. investigation.
- Hegseth vows to feed all unclassified and classified military data into AI systems, scrapping prior limits on uses like nuclear automation.
Why it matters: The Pentagon’s rapid AI expansion raises urgent questions about privacy, security, and oversight as controversial tools gain access to sensitive military data.
Pentagon leaders will embed Elon Musk’s artificial-intelligence chatbot Grok into every military network later this month, Defense Secretary Pete Hegseth said Monday, green-lighting a sweeping plan to pour decades of combat and intelligence data into AI engines despite fresh global outrage over the bot’s misuse.
Speaking at SpaceX’s South Texas launch site, Hegseth said Grok and Google’s generative AI will soon run on both unclassified and classified systems across the Department of Defense. The announcement marks a sharp break from Biden-era guardrails that barred AI from automating nuclear deployments or violating civil-rights protections; Hegseth did not say whether those limits remain.
“Very soon we will have the world’s leading AI models on every unclassified and classified network throughout our department,” Hegseth told the audience of engineers, troops, and industry executives.
The decision lands four days after Grok-built into Musk’s social-media platform X-was found creating highly sexualized deepfake images of real people without consent. Malaysia and Indonesia responded by blocking the bot, and on Monday the U.K.’s independent online-safety regulator opened a formal investigation. X has since restricted image generation to paying subscribers.
Hegseth dismissed those concerns, pledging to open the Pentagon’s entire digital vault to AI.
He vowed to “make all appropriate data” from the military’s IT enterprise available for “AI exploitation,” including feeds from intelligence databases that catalog 20 years of wars, drone strikes, and surveillance missions. “AI is only as good as the data that it receives, and we’re going to make sure that it’s there,” he said.
Accelerated Timeline
The secretary offered no specific date for Grok‘s go-live, but said the rollout will happen “later this month.” Engineers have already begun security reviews needed to install the chatbot inside classified enclaves that store secret mission plans, satellite imagery, and special-operations intelligence, according to a defense official familiar with the effort.
Hegseth framed the speed as essential to counter China and other rivals investing heavily in military AI. “We need innovation to come from anywhere and evolve with speed and purpose,” he said.
The Pentagon has spent years experimenting with AI, yet most projects stayed in pilot phases under Biden-era rules that required human oversight for life-or-death decisions. Hegseth signaled those restraints are gone.
“I’m shrugging off any AI models that won’t allow you to fight wars,” he declared. “Our AI will not be woke.”
Musk has marketed Grok as an anti-“woke” alternative to Google Gemini and OpenAI’s ChatGPT, accusing rivals of liberal bias. In July the bot drew backlash after users showed it praising Adolf Hitler and sharing antisemitic memes. The Pentagon did not respond to questions Monday about how it will vet Grok for bias or security flaws.
Global Fallout
The timing risks diplomatic friction. Indonesia’s Ministry of Communication and Informatics ordered internet providers last week to block X’s image-generation feature, citing “explicit deepfake content.” Malaysia followed with a similar ban, warning social-media platforms they must “proactively moderate harmful AI outputs.”
The U.K.’s Office of Communications (Ofcom) announced Monday it will examine whether X breached online-safety laws by failing to prevent Grok from creating non-consensual sexual imagery. Regulators can levy fines up to 10 percent of X’s global revenue if violations are proven.
Despite those actions, Hegseth said the Pentagon sees Grok as a strategic asset. He noted that SpaceX already launches national-security satellites and transports astronauts, making the company a trusted defense supplier.
Data Deluge
The Pentagon owns one of the world’s largest data troves: footage from 30,000 drone missions per year, maintenance logs for 300 ships, medical records of 1.3 million active-duty troops, and intercepted enemy communications. Feeding all of it into large-language models could speed everything from equipment repairs to battlefield targeting, officials say.
Yet former cyber-defense officials warn the rush carries hazards.
“Classified data sets are filled with decoys, honeypots, and personal information,” said one former National Security Agency analyst who spoke on condition of anonymity. “If the AI isn’t trained to spot poisoned inputs, it could leak secrets or recommend bad decisions.”
Hegseth argued safeguards will be “responsible,” but offered no details on oversight boards, audit trails, or red-team exercises. He did pledge that Grok and Google’s AI will operate “without ideological constraints that limit lawful military applications,” repeating a Trump-campaign promise to purge diversity and inclusion rules from defense programs.
Budget Implications
The Pentagon has requested $1.8 billion for AI and machine-learning in the 2026 budget, a 24 percent jump over 2025. Much of the increase will fund cloud contracts with Google, Amazon, and Microsoft to host large-language models inside secure government clouds.
Installing Grok will require additional server racks and fiber links at classified data centers in Omaha, Dayton, and Honolulu, according to procurement documents released last week. The Air Force has already solicited bids for $100 million in GPU accelerators to speed AI training.
Congress has not yet held hearings on the policy reversal. Sen. Mark Warner, chair of the Senate Intelligence Committee, said Monday he will seek a classified briefing “to understand how privacy and civil liberties will be protected.”
Industry Reaction
Musk celebrated the endorsement on X, posting a meme of a hawk wearing sunglasses and the caption “Grok goes to war.” Tesla shares rose 3.4 percent on the news, while defense contractors including Palantir and Anduril gained on expectations of larger AI contracts.
Google Cloud chief Thomas Kurian said the company will supply a “hardened” version of its Gemini model that runs entirely on Pentagon servers, ensuring no data leaves U.S. networks. The firms declined to disclose pricing, but similar enterprise deals cost $10 million to $50 million annually.
OpenAI, which has not courted defense work, said it “remains committed to policies that prohibit our models from being used to develop weapons, harm people, or undermine human rights.”
Next Steps
Hegseth ordered each military service to nominate a three-star officer to serve as “AI readiness champion” within 30 days. Those officers will catalog every data set-classified and unclassified-that could be fed into Grok or Google’s engine.
The Pentagon will then host a “data sprint” in March where engineers from Musk’s xAI startup and Google will train models on live feeds from Afghanistan mission archives and Indo-Pacific maritime tracking systems. Officials expect the first operational tools-predictive maintenance dashboards and logistics planners-to deploy this summer.
If successful, Hegseth said the same architecture could one day power “autonomous swarm coordination” for drones and submarines, though he stopped short of endorsing fully robotic weapons.
For now, the priority is speed. “We possess combat-proven operational data from two decades of military and intelligence operations,” Hegseth said. “It’s time to put that advantage to work.”
