MarketCrystal | February 28, 2026
When we built MarketCrystal, one decision was non-negotiable: Claude, built by Anthropic, was the only AI we trusted to power this platform. Not OpenAI. Not Google Gemini. Not Grok. That stance didn’t come from brand loyalty — it came from a documented, public record of how each company handles the most important question in AI right now: what are your limits, and will you hold them under pressure?
Yesterday, we got the clearest answer to that question we’ve ever seen.
Here’s what happened — no opinion, just the documented record.
The Two Red Lines
Anthropic had a $200 million contract with the Pentagon to run Claude on classified military networks — the only AI model approved for that use. Inside that contract were two restrictions Anthropic required of all customers, including the government:
- Claude will not be used for mass domestic surveillance of Americans
- Claude will not be used to power fully autonomous weapons systems
Anthropic’s CEO Dario Amodei stated publicly: “Using these systems for mass domestic surveillance is incompatible with democratic values” and that “frontier AI systems are simply not reliable enough to power fully autonomous weapons.”
Those were the two red lines. Two. Out of every possible military application, those were the only ones Anthropic would not permit.
The Pentagon’s Demand
The Pentagon’s position: it must have access to any AI it contracts for “all lawful purposes” — no private company gets to set limits on how the military uses technology it licenses.
Pentagon officials stated publicly that they had no intention of using AI for mass surveillance or autonomous weapons. But they also refused to put that in writing as a contractual guarantee.
Defense Undersecretary Emil Michael set a deadline: 5:01 PM ET, Friday February 27, 2026. Drop the two restrictions, or lose the contract.
Anthropic’s response: “Threats do not change our position: we cannot in good conscience accede to their request.”
The deadline passed without agreement.
What Happened Next
Within hours of the deadline:
President Trump posted on Truth Social directing “EVERY Federal Agency in the United States Government to IMMEDIATELY CEASE all use of Anthropic’s technology.”
Defense Secretary Pete Hegseth designated Anthropic a “Supply-Chain Risk to National Security” — a designation previously applied to foreign adversaries like Chinese telecom giant Huawei. The designation bars any military contractor, supplier, or partner from conducting any commercial business with Anthropic, extending the blacklist beyond government into the private sector.
Trump also warned Anthropic to cooperate during a six-month phaseout period or face “the Full Power of the Presidency” with “major civil and criminal consequences to follow.”
Senator Mark Warner, Vice Chair of the Senate Intelligence Committee, responded: “President Trump and Secretary Hegseth’s efforts to intimidate and disparage a leading American company — potentially as the pretext to steer contracts to a preferred vendor whose model a number of federal agencies have already identified as a reliability, safety, and security threat — pose an enormous risk.”
What the Other AI Companies Did
OpenAI — Hours after the Trump announcement, CEO Sam Altman posted that OpenAI had struck a deal with the Pentagon to deploy its models on classified networks. In a memo to his own employees the previous day — before the deal — Altman had written: “We have long believed that AI should not be used for mass surveillance or autonomous lethal weapons, and that humans should remain in the loop for high-stakes automated decisions. These are our main red lines.” He also wrote that “this is no longer just an issue between Anthropic and the DoW; this is an issue for the whole industry.” OpenAI then signed the Pentagon deal.
xAI (Grok) — Already approved for classified Pentagon use. Signed on without public restrictions.
Google, Meta — Pentagon contracts in place. No public restrictions stated. Over 100 Google employees sent an internal letter to management asking for the same limits Anthropic demanded. No public response from leadership.
Anthropic — Stated it would challenge the supply-chain risk designation in court. Noted under federal law the designation can only apply to Pentagon contracts, not Anthropic’s broader commercial business. Said it had “tried in good faith” over months of negotiations and that its two restrictions “have not affected a single government mission to date.”
The Documented Timeline
- July 2025 — Anthropic signs $200M Pentagon contract. Claude becomes the only AI on classified military networks.
- Months of negotiations — Pentagon presses Anthropic to remove restrictions. Anthropic refuses on both points.
- Tuesday, Feb 24 — High-stakes meeting between Hegseth and Amodei at the Pentagon.
- Thursday, Feb 27 — Anthropic publicly states it will not remove restrictions. Simultaneously, Emil Michael offers Anthropic a deal via phone — one that would have “required allowing the collection or analysis of data on Americans, from geolocation to web browsing data to personal financial information.”
- 5:01 PM ET, Friday Feb 27 — Deadline passes. Trump and Hegseth announce blacklist.
- Friday evening — OpenAI announces Pentagon deal for classified networks.
Why This Matters for MarketCrystal
This platform exists because we believe data without integrity is worthless. The same principle applies to AI. We chose Anthropic because they publish their principles, build them into their products contractually, and — as of yesterday — held them under the most extreme commercial pressure a company can face.
The supply-chain risk designation could cost Anthropic a significant portion of their enterprise customer base. Any company with a Pentagon contract now has to prove they don’t use Claude. That’s an existential business threat designed to force compliance.
They didn’t comply.
That’s the record. You can draw your own conclusions.
MarketCrystal provides trend analysis for informational purposes only. This is not financial advice. All quotes sourced from Axios, NPR, CNN, NBC News, CNBC, Fortune, The Hill, and Federal News Network reporting dated February 27–28, 2026.