Pentagon Eyes Hacking AI Models for Cyber Operations Despite Anthropic Supply-Risk Ban
Updated
Updated · Gizmodo · May 21
Pentagon Eyes Hacking AI Models for Cyber Operations Despite Anthropic Supply-Risk Ban
8 articles · Updated · Gizmodo · May 21
Politico reported the Pentagon is exploring frontier AI models with hacking capabilities—including Anthropic’s unreleased Claude Mythos Preview—for both defensive and offensive cyber missions.
The push expands a task force announced earlier this month by Cyber Command and the NSA, which was initially framed around safely deploying leading AI on highly classified “high-side” systems.
Charles Moore, a former Cyber Command deputy commander, said such tools can detect threats, prioritize vulnerabilities, speed decisions and conduct cyber operations faster than adversaries.
Anthropic’s role is complicated: the Pentagon has formally labeled the company a supply-chain risk, a designation that in theory bars both the department and its contractors from working with it.
The reported interest still fits a broader national-security push around Mythos, with the White House already involved in Project Glasswing’s rollout and reportedly holding veto power over some access expansions.
As military AI learns to hack, what prevents these digital weapons from turning on their creators or escaping control?
In the AI arms race, how can new models avoid triggering an accidental global conflict through battlefield miscalculation?
U.S. Military AI in 2026: Strategic Partnerships, Anthropic’s Exclusion, and the Ethics of Warfighting Algorithms
Overview
In May 2026, the U.S. Department of Defense launched new partnerships with eight top AI companies to boost military capabilities and maintain technological superiority. This move, driven by a mandate from President Trump and Secretary Hegseth, aims to deploy advanced AI across classified Pentagon systems and counter emerging threats. Notably, Anthropic was excluded after a dispute over ethical restrictions and operational control, leading to its designation as a 'supply chain risk.' The Pentagon’s strategy highlights the balance between rapid AI adoption, national security needs, and the challenges of aligning government and industry standards for responsible AI use.