Cloudflare stock jolted after Anthropic security move

Cloudflare did not announce bad earnings this week, but its stock still got hit as investors tried to price a new threat in real time: Anthropic launched tools that move more of the artificial intelligence stack onto Anthropic’s own platform, and Cloudflare shares were swept into the selloff. On April 8, Anthropic launched Claude Managed Agents, which it described as cloud-hosted agents built from composable application programming interfaces, and on April 7 it launched Project Glasswing, which gives a select group access to Claude Mythos Preview for defensive cybersecurity work. A managed agent is basically an artificial intelligence worker that runs on someone else’s servers instead of your own, like renting a fully staffed kitchen instead of buying ovens, hiring cooks, and wiring the building yourself. Anthropic said its service handles sandboxed code execution, checkpointing, credential management, scoped permissions, and tracing for developers. That matters to Cloudflare because Cloudflare sells the plumbing around internet traffic, application security, and developer infrastructure, and it has spent the past year pitching itself as a place to build and secure artificial intelligence applications. Cloudflare’s AI Security Suite says customers can discover unsanctioned tools, protect models from abuse, secure agent access, and prevent prompt data leaks. Cloudflare also worked directly with Anthropic before this selloff. In May 2025, Cloudflare said companies including Asana, Atlassian, PayPal, Sentry, Stripe, and Webflow were using secure connections built on Cloudflare Workers so Claude could interact with their services on behalf of users. Then the relationship changed shape. Instead of just being a model company that needed outside infrastructure partners, Anthropic started offering more of the surrounding machinery itself, first with tightly controlled cyber tools in Project Glasswing and then with hosted agent infrastructure in Claude Managed Agents. Wall Street has been jumpy about this exact idea for weeks. On March 27, CNBC reported that cybersecurity stocks slumped after a report that Anthropic was testing Mythos, and on April 7 Reuters reported that Anthropic said Mythos Preview had already found “thousands” of major vulnerabilities in operating systems, web browsers, and other software. Anthropic did not throw Cloudflare out of Project Glasswing. Anthropic listed Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks as launch partners, and said it extended access to more than 40 additional organizations that build or maintain critical software infrastructure. That partner list may be part of why the market reacted so hard. If Anthropic can put Apple, Microsoft, Amazon Web Services, CrowdStrike, and Palo Alto Networks inside the tent while keeping Mythos Preview restricted, investors start asking which vendors gain a head start and which vendors get squeezed on pricing or relevance. Cloudflare’s drop was also bigger than one headline. Yahoo Finance said the April 10 move was amplified by a broader software selloff after a UBS downgrade of ServiceNow, while CNBC’s quote page showed Cloudflare trading down from a previous close of $193.05 to an intraday low of $164.05. The cleanest way to read the move is that investors are no longer valuing artificial intelligence vendors as “just model makers.” When Anthropic ships the model, the agent runtime, and a restricted security program in the same week, companies like Cloudflare get judged against a bigger rival footprint than they were a month ago.