Pentagon Deadline Looms for AI Firm Anthropic Over Military Use Policy

Anthropic faces a Pentagon deadline over military AI use, risks a $200M contract. It also acquires Vercept to boost autonomous capabilities while revising its safety policy.

Anthropic, the artificial intelligence company, faces a critical decision point with the U.S. Department of Defense. A deadline set by the Pentagon expires today at 5:01 PM Eastern Time, with a lucrative government pilot program hanging in the balance. The standoff centers on CEO Dario Amodei’s continued refusal to lift restrictions on the military applications of the company's AI models.

Acquisition and Technological Advancements

On the technology front, Anthropic is moving to enhance its software's autonomous capabilities through the acquisition of Seattle-based startup Vercept. The nine-person team will be integrated following the recent update to the Claude Sonnet 4.6 model, which dramatically improved its performance on operating system benchmarks from under 15 percent to 72.5 percent.

Vercept was last valued at $67 million in January 2025. Its standalone desktop application will be discontinued within the next 30 days. The market reacted swiftly to the progress in autonomous computer capabilities: shares of software provider UiPath declined by 3.6 percent following the acquisition announcement.

Core Conflict: Ethical Principles vs. Government Demands

The dispute originates from a demand by Defense Secretary Pete Hegseth that Anthropic make its software available for all lawful military purposes. The firm has drawn a line, refusing to allow its technology to be used for autonomous weapon systems or mass domestic surveillance. Failure to reach an agreement before the deadline risks the termination of a contract worth up to $200 million.

Furthermore, the Defense Department is considering classifying Anthropic as a supply chain risk or compelling cooperation through the Defense Production Act. The company's stance contrasts with competitors like xAI, which have already agreed to terms for deployment within classified networks. The central question remains whether Anthropic will compromise its ethical guidelines for the Pentagon contract.

Strategic Pivot in Safety Policy

Amid the government confrontation, Anthropic’s management is concurrently overhauling its operational safety framework. The company is relaxing its 2023 "Responsible Scaling Policy." Under the revised approach, the development of new models will no longer be halted solely due to theoretical risks while rivals continue to advance.

Instead of implementing development pauses, Anthropic will now rely on detailed risk assessments published every three to six months. This shift is a direct response to mounting competitive pressures within the rapidly evolving AI sector.

Should no agreement be reached by this evening, Anthropic stands to lose not only the $200 million deal but also risks a direct confrontation with the U.S. government, potentially triggering far-reaching regulatory consequences.