Anthropic under pressure on multiple fronts

Anthropic is facing simultaneous product, legal and security headwinds: its Mythos model raised cybersecurity concerns that prompted high-level talks involving Treasury and the Fed, the company is pitching managed agents to enterprise customers, and a federal court recently denied its request to pause a U.S. 'supply chain risk' label. Those developments together tighten Anthropic’s operating environment for both sales to regulated customers and public-stakeholder relations. (x.com) (the-decoder.com) (euronews.com)

Anthropic is getting squeezed from three directions at once: Washington is treating one of its newest models as a cyber risk, a federal appeals court just refused to pause a damaging government label, and the company is simultaneously trying to sell a new “managed agents” product to big corporate buyers. (cnbc.com) (politico.com) (the-decoder.com) The sharpest jolt came this week, when Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell met major United States bank chiefs to discuss the cyber threat from Anthropic’s new Mythos model. Anthropic had already limited Mythos access instead of releasing it broadly. (cnbc.com) (anthropic.com) Anthropic’s own write-up says Mythos Preview can identify and exploit zero-day vulnerabilities in every major operating system and every major web browser when a user directs it to do so. The company said it does not plan to make Mythos Preview generally available and is first channeling it through a defensive security effort called Project Glasswing. (anthropic.com 1) (anthropic.com 2) Project Glasswing is Anthropic’s attempt to keep a very sharp tool inside a guarded workshop. Anthropic said launch partners include Amazon Web Services, Apple, Cisco, Google, JPMorganChase, Microsoft, Nvidia, Palo Alto Networks, and the Linux Foundation, plus more than 40 additional organizations that maintain critical software infrastructure. (anthropic.com 1) (anthropic.com 2) At almost the same moment, Anthropic rolled out Claude Managed Agents in public beta, which lets companies run autonomous software agents on Anthropic’s own infrastructure instead of building the plumbing themselves. Anthropic’s engineering team says the service packages three pieces together: a session log, a harness that routes tool calls, and a sandbox where the model can run code and edit files. (the-decoder.com) (anthropic.com) That product pitch is aimed straight at enterprise buyers who want agents without operating their own control room. Anthropic’s documentation says Managed Agents is available through the Claude application programming interface, and The Decoder reported early users including Notion, Rakuten, and Sentry, with pricing at $0.08 per session hour on top of token charges. (anthropic.com) (the-decoder.com) The legal pressure comes from a separate fight with the Pentagon. On April 8, a three-judge panel on the District of Columbia Circuit denied Anthropic’s request to pause a Department of Defense designation that labels the company a “supply chain risk” while the appeal continues. (politico.com) (cnbc.com) The court said it would move quickly because Anthropic could suffer irreparable harm, but it still left the label in place for now. CNBC reported that the Defense Department had imposed the designation in early March and said Anthropic threatened national security. (bloomberg.com) (cnbc.com) That creates an awkward overlap between the company’s sales story and its political reality. Anthropic is asking corporations to trust it with long-running autonomous agents inside sensitive workflows at the same time that top financial regulators are warning banks about one Anthropic model and the Pentagon is still fighting over whether the company belongs in defense systems at all. (the-decoder.com) (cnbc.com) (politico.com) Anthropic’s problem is not one bad headline but three different audiences reacting at once. Regulators are looking at systemic cyber risk, procurement officials are looking at national security labels, and enterprise customers are being asked to move more of their software work onto Anthropic-managed systems. (cnbc.com 1) (cnbc.com 2) (anthropic.com) If Anthropic can keep Mythos boxed into defensive use, win back ground in court, and persuade companies that managed agents are safer on its platform than on their own servers, this week will look like a rough transition. If it cannot, April 2026 may be remembered as the week its fastest product push collided with the limits of public trust. (anthropic.com) (bloomberg.com) (the-decoder.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.