Key Insights:
- Polymarket odds on the Pentagon banning Claude jumped 30%, from 15-27% to 49%, after the CEO’s rejection.
- Anthropic refused Pentagon demands for mass surveillance and autonomous weapons using Claude AI.
- Polymarket odds crashed 36 points from a 49% peak to just 13%, showing traders doubt the ban will happen.
Polymarket odds jumped on Feb. 26, 2026. The market asked if Pete Hegseth would ban Claude by March 31. Before Anthropic CEO Dario Amodei rejected Pentagon demands, odds sat between 15% and 27%. After his rejection, the odds spiked to 49%. That’s a 30% jump. Traders saw the refusal and immediately priced in higher ban risk.
The Pentagon gave Anthropic a final offer with a 24-hour deadline. Remove guardrails for mass surveillance of Americans and allow Claude to use autonomous lethal weapons. Amodei said no. “These threats do not change our position. We cannot in good conscience accede.” Polymarket reacted in real time.
Anthropic CEO Rejects Pentagon Surveillance Demands
The Pentagon awarded Anthropic approximately $200 million in July 2025 for AI national security tools. But they wanted more access than Anthropic’s safety guidelines allow. They demanded guardrail removal for mass surveillance of Americans and autonomous weapons deployment.
Dario Amodei released a public letter on Feb. 26, 2026, explaining the refusal. He wrote, On autonomous weapons, Amodei said Claude is not technically capable of safe deployment. The AI could make life-or-death mistakes.
The Pentagon responded with threats. Secretary of Defense Pete Hegseth could ban Claude entirely. The Defense Department could invoke the Defense Production Act, forcing compliance. Anthropic could be blacklisted from defense contracts.
Other AI companies reportedly complied. Sources suggest xAI, OpenAI, and Google agreed to similar Pentagon requests. Anthropic stood alone, refusing. Amodei offered to help transition the Pentagon to another provider.
Polymarket Odds Jump 30% to 49% on Ban Question in Hours
The Polymarket market launched around Feb. 16, 2026. “Will Pete Hegseth ban Claude by March 31?” The market resolves YES if DoD ends its Anthropic relationship or bans Claude’s use by March 31. Otherwise, no.
Early odds were 15% to 27%. Most traders thought the Pentagon needed Claude too much to ban it. Amodei’s February 26 rejection changed that calculation.
Within hours, YES odds surged to 49%. Almost a coin flip. Traders suddenly saw the ban as equally likely. The 30% spike showed how much weight the market put on Anthropic’s refusal.
Trading volume reached $303,000 to $318,000. Significant for a niche AI policy question. Liquidity stayed moderate, whereas some users complained that large sell orders caused temporary price swings.
The Polymarket market became a real-time oracle. Traditional media moved more slowly. Traders with money at risk processed news instantly.
Polymarket Cuts Ban Odds on Claude to 36%
The 49% spike didn’t last. Within hours, the odds corrected to 34%. Still higher than pre-rejection levels but far below the panic spike. At press time, the number sits at 13%, down over 70% since those staggeringly high ban hope numbers.
Traders reassessed the situation. The Pentagon needs advanced AI. Claude ranks among the top models. Anthropic leads in some safety benchmarks. Banning Claude means losing access to one of the best systems.
The Polymarket correction suggests traders believe the Pentagon is bluffing. Threats are cheap. Actually, banning is costly. If the military needed unrestricted AI, they could use models from companies that already agreed.
Some analysis pointed to precedent. If the government forces Anthropic to comply via the Defense Production Act, other AI companies lose power. But if Anthropic resists successfully, it sets limits on government control.
For now, Polymarket prices an 87% chance that the ban doesn’t happen. Traders think Anthropic and the Pentagon will find a compromise. Or the Pentagon will quietly drop demands rather than lose access. The market bets on pragmatism now.