NVIDIA Corp., Microsoft Corp., Reflection AI Inc., and Amazon.com Inc. have all recently struck agreements with the Department of Defense to integrate advanced artificial intelligence (AI) into the nation’s most sensitive classified networks.
The deals cement a defining pivot toward what officials call an “AI-first fighting force” while underscoring tensions with Anthropic. The military intends to leverage these latest tools to streamline data analysis, sharpen situational awareness, and augment warfighter decision-making in increasingly volatile global environments.
These companies join SpaceX, OpenAI, and Google, which recently agreed to provide AI tools for Pentagon classified networks. The statement marks the first official Pentagon confirmation of a new deal with Google that was reported earlier this week.
While giants like Microsoft and Amazon have long-standing ties to the Pentagon, the inclusion of newcomers like Elon Musk’s SpaceX and the specialized firm Reflection AI highlights the military’s urgency to capture cutting-edge generative and analytical capabilities. According to Defense Department statements, these partnerships will allow for the “lawful operational use” of AI systems within highly restricted environments.
“Given all the market rhetoric, it is hardly surprising that the Pentagon and DoW (Department of War) is moving on. What will be interesting is the long-term impact on Anthropic,” Steven Dickens, CEO of HyperFRAME Research, said in an email. “Will the lost revenue with DoW and the Pentagon translate into more business with enterprises, or will it drag on growth with basically a sector lost? Also, will this drag on revenue also manifest in other five eyes and NATO nations?”
Pentagon officials, speaking on the condition of anonymity, confirmed that the agreement with Amazon Web Services was finalized late Thursday evening. The broader coalition aims to ensure that the U.S. maintains a technological edge over adversaries by embedding autonomous capabilities directly into the military’s digital backbone.
As the Pentagon doles out contracts to some of tech’s biggest names, it continues to feud with Anthropic. Despite a $200 million deal to handle classified materials, the San Francisco-based firm has been declared a “supply-chain risk” and effectively banned from federal networks.
The rift reportedly stems from Anthropic’s refusal to relax its internal red lines regarding mass domestic surveillance and fully autonomous weaponry. In retaliation, the federal government moved to bar the company’s products, leading to a legal battle in which Anthropic recently secured a temporary injunction.
Tensions were further exacerbated by the Trump administration’s intervention this week to halt expansion of Anthropic’s Mythos model. Developed under the secretive Project Glasswing, Mythos is a breakthrough in autonomous cybersecurity capable of identifying and exploiting zero-day vulnerabilities, flaws unknown even to the software’s original developers.
While Anthropic sought to expand Mythos access from 50 to 120 organizations to bolster national infrastructure, the White House formally opposed the move. Officials cited “grave national security risks,” fearing the model’s ability to autonomously dismantle digital defenses could be catastrophic if misappropriated.
Emil Michael, the Pentagon’s chief technology officer, noted that while the model is a “national security moment” for its patching capabilities, the government remains wary. “We have to make sure our networks are hardened,” Michael said, “because that model has capabilities that are particular to finding cyber vulnerabilities.”
As the Pentagon pursues its “AI-first” doctrine with its seven chosen partners, the standoff with Anthropic serves as a stark reminder of the ethical and security friction points defining the new era of algorithmic warfare.
