The military wants more options, not fewer, as AI becomes embedded in operations.
In the ongoing effort to embed artificial intelligence into the sinews of modern warfare, the Pentagon has formalized partnerships with seven of the world's most powerful technology companies, opening its most classified networks to their systems. The deliberate exclusion of Anthropic — designated a supply-chain risk after resisting Defense Department restrictions — speaks to a deeper tension between the autonomy of AI developers and the imperatives of national security. The military's embrace of a multi-vendor architecture reflects a hard-won institutional wisdom: that dependence on any single intelligence, artificial or otherwise, is its own vulnerability. The question of whether Anthropic will eventually find its way back into this fold remains, for now, a matter of politics as much as technology.
- The Pentagon has opened its most sensitive classified networks — Impact Levels 6 and 7 — to AI systems from OpenAI, Google, Microsoft, Amazon, NVIDIA, SpaceX, and Oracle, marking the military's most aggressive integration of AI into warfare planning and targeting.
- Anthropic's absence is conspicuous and deliberate: the company was designated a supply-chain risk two months ago after resisting Pentagon restrictions, and its tools are now being phased out across the department and its contractors within six months.
- Despite internal familiarity with Anthropic's capabilities — Pentagon staff and contractors considered them competitive — institutional orders have overridden preference, illustrating how quickly a compliance dispute can sever a high-value technology relationship.
- The Pentagon's GenAI.mil platform has already reached 1.3 million Defense Department personnel in just five months, a scale that underscores how rapidly AI has become load-bearing infrastructure in military operations.
- By deliberately spreading contracts across multiple vendors, the Pentagon is guarding against 'vendor lock' — a strategic acknowledgment that over-reliance on any single AI provider poses its own operational risk.
- Anthropic's exclusion may not be permanent: President Trump has suggested the company is 'shaping up,' and the Pentagon's CTO has separately flagged Anthropic's Mythos model as a distinct concern, leaving the door open to a future reconciliation.
On a Friday in early May, the Pentagon announced agreements with seven major AI companies — OpenAI, Google, Microsoft, Amazon Web Services, NVIDIA, SpaceX, and Oracle — to deploy their systems on the military's most classified networks. One prominent name was missing: Anthropic, which had resisted Pentagon restrictions on how its tools could be used and was designated a supply-chain risk two months prior.
The exclusion carried weight. Pentagon staff and defense contractors had grown accustomed to Anthropic's capabilities and considered them competitive with alternatives, but orders were already in place to phase the company's tools out over the next six months. The military was signaling a deliberate pivot.
The Pentagon's goal was to give troops faster access to AI for planning, logistics, and targeting across all domains of warfare, embedding these tools into Impact Levels 6 and 7 — the networks where the most sensitive operations are handled. The scale of existing adoption made the stakes clear: GenAI.mil had already reached more than 1.3 million Defense Department personnel within five months of launch.
By partnering with multiple vendors rather than consolidating around one, the Pentagon was explicitly guarding against "vendor lock" — a recognition of how deeply AI had become woven into military infrastructure. The move also created openings for smaller defense technology startups in a space long dominated by established players.
Anthropics's exclusion may not be final. Pentagon CTO Emil Michael flagged the company's Mythos model as a separate concern — a system that had raised alarms for potentially accelerating cyberattacks — while President Trump recently suggested Anthropic was "shaping up" in his administration's view. Whether the company eventually rejoins the Pentagon's roster remains an open question, shaped as much by politics as by technology.
On a Friday in early May, the Pentagon announced it had struck agreements with seven major artificial intelligence companies to deploy their systems on the military's most sensitive networks. The list—OpenAI, Google, Microsoft, Amazon Web Services, NVIDIA, SpaceX, and Oracle—represents the Defense Department's most expansive push yet to embed AI into classified operations. One prominent name was absent: Anthropic, the AI startup that had resisted Pentagon restrictions on how its tools could be used.
The timing was pointed. Just two months earlier, the Pentagon had designated Anthropic a supply-chain risk and begun blocking its software from use across the department and its contractors. Now, with this new slate of partnerships, the military was signaling a deliberate pivot away. Pentagon staff, former officials, and defense contractors told Reuters they had grown accustomed to Anthropic's capabilities and considered them competitive with alternatives. But orders were already in place to phase the company's tools out over the next six months.
The Pentagon's stated goal was straightforward: give troops faster, broader access to AI for planning, logistics, and targeting decisions across all domains of warfare. The new agreements would bring these tools into Impact Levels 6 and 7—the classified networks where the most sensitive military operations are handled. The Defense Department framed the move as essential to establishing what it called an "AI-first fighting force."
What made the expansion significant was its scale and the principle behind it. The Pentagon's internal platform, GenAI.mil, had already reached more than 1.3 million Defense Department personnel within five months of launch—adoption rates that would be remarkable for any enterprise software, let alone one operating in a classified environment. By bringing in multiple vendors rather than consolidating around a single provider, the Pentagon was explicitly trying to avoid what it called "vendor lock," a recognition of how deeply AI had become woven into military infrastructure.
The exclusion of Anthropic, however, was not necessarily permanent. Emil Michael, the Pentagon's Chief Technology Officer, acknowledged that Anthropic still presented supply-chain concerns, but he also singled out the company's Mythos model as a separate issue—a system that had raised alarms across government and industry for potentially accelerating cyberattacks. Access to Mythos had been extended to select organizations to help strengthen defenses, though it remained unclear whether the Pentagon was among them. President Trump had recently suggested that Anthropic was "shaping up" in his administration's view, hinting that the current standoff might not endure.
For now, the Pentagon was moving forward with a broader roster of AI partners. The shift created new openings for smaller defense technology startups seeking to break into government contracts, a space traditionally dominated by a handful of established players. The military's message was unmistakable: as AI became embedded in operations, it wanted options, not dependence. Whether that calculus would eventually include Anthropic again remained an open question.
Citações Notáveis
The Pentagon said the agreements would accelerate transformation toward establishing the U.S. military as an AI-first fighting force and strengthen warfighters' ability to maintain decision superiority across all domains of warfare.— Pentagon statement
Pentagon staff and contractors told Reuters they had grown accustomed to Anthropic's tools and considered them stronger than many alternatives.— Reuters reporting
A Conversa do Hearth Outra perspectiva sobre a história
Why did the Pentagon suddenly exclude Anthropic when it had been working with them?
It wasn't sudden—there was a conflict two months before. Anthropic pushed back on restrictions the Pentagon wanted to place on how its AI could be used. The Pentagon responded by labeling them a supply-chain risk, which is a formal way of saying they don't trust the company in sensitive systems.
But people inside the Pentagon actually liked Anthropic's tools, right?
Yes. Pentagon staff and contractors told Reuters they preferred Anthropic's capabilities to many of the alternatives. That's what makes the exclusion notable—it's not about technical inferiority. It's about control and trust.
So this announcement with seven other companies is partly about sending a message?
Absolutely. The Pentagon is saying: we don't need you, and we have plenty of other options. It's also about avoiding what they call vendor lock-in—not depending on any single company for critical military AI.
Is the door completely closed on Anthropic?
No. The Pentagon's CTO said Anthropic still presents concerns, but he also suggested the situation might not be permanent. Trump recently said Anthropic was "shaping up," which hints the standoff could end.
What's the Mythos model he mentioned?
It's an Anthropic system that's raised alarms because it could potentially accelerate cyberattacks. It's being shared with select organizations to help them strengthen defenses, but it's unclear if the Pentagon is one of them.
So the real story is about power and leverage in military AI?
Exactly. The Pentagon wants multiple vendors so no single company can dictate terms. Anthropic learned that lesson the hard way.