The Dual Reality of Anthropic: Offensive Cyber and Global Moratoriums

Anthropic is simultaneously embedding engineers within the NSA to deploy offensive AI and calling for a global pause on AI development. Reports indicate the company has placed roughly half a dozen engineers at the NSA to customize its Mythos AI model for offensive cyber operations. These operations could potentially include attacks on networks in countries like China and Iran.

The deployment of Mythos represents a strategic contradiction. While Anthropic has declined to release the model publicly due to misuse risk, the company is actively customizing it for state-sponsored cyber warfare. This model is currently accessible to a restricted coalition known as Project Glasswing, which includes Microsoft, Apple, and Amazon.

The tension between Anthropic’s commercial ambitions and its security stance is deepening. The company is currently suing the Pentagon after Defense Secretary Pete Hegseth designated Anthropic a supply-chain risk. This designation followed the collapse of a $200 million contract, triggered by Anthropic's refusal to allow the DoD to use Claude for domestic mass surveillance or fully autonomous weapons. While a California judge blocked the blacklisting as apparent First Amendment retaliation, the NSA contract remained exempt from such bans, allowing the agency to continue using Mythos.

This friction occurs as Anthropic files for an IPO that could value the company above $1 trillion.

The company's internal research also highlights a shift toward recursive self-improvement. Anthropic’s research institute recently published findings on how Claude is automating its own development. The data shows that Claude now writes more than 80% of the code merged into Anthropic's production codebase. This is a significant jump from the low single digits seen before Claude Code launched in early 2025. Consequently, engineers are shipping roughly eight times as much code per day as they did in 2024.

Anthropic is using this momentum to argue for a global moratorium on the AI arms race, suggesting a framework similar to Cold War-era nuclear treaties.

The company is effectively building the tools for the next generation of cyber warfare while simultaneously lobbying for the brakes to be applied to the industry. This creates a massive gap between the capabilities being deployed for national security and the regulatory frameworks being proposed for the public.

Watch how the litigation regarding the Pentagon's supply-chain designation progresses, as the outcome will define the boundaries of private AI companies acting as state-level intelligence partners.