Anthropic move rattles Cloudflare

Anthropic’s new cybersecurity push jolted software stocks this week, and Cloudflare closed Friday down 13.5% after investors reassessed how much AI could encroach on security vendors. (anthropic.com) (stockanalysis.com) Anthropic said on April 7 that it is limiting access to Claude Mythos Preview, a model it says is especially strong at finding software flaws, and is rolling it out through a new program called Project Glasswing. Anthropic listed Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, Nvidia and Palo Alto Networks among the launch partners, plus more than 40 additional organizations. (anthropic.com) (cnbc.com) The basic idea is simple: software bugs are hidden mistakes in code, and Anthropic says its model can help defenders spot and fix them before attackers do. Anthropic also said it is restricting the model because the same skill that finds a weakness can also be used to exploit one. (anthropic.com) (cnbc.com) That landed in a market already primed to treat AI as both a product feature and a threat to existing software businesses. Reuters- and Barron’s-cited reports described a broad selloff on April 9, with Cloudflare down 8.6% that day, Snowflake down 11.8%, Zscaler down 11.4% and CrowdStrike down 7.5%. (crn.com) Cloudflare was an obvious name to get pulled into that reaction because it has been pitching AI security as a growth area of its own. On March 19, 2025, the company launched “Cloudflare for AI,” including tools to discover AI use inside a company and block threats such as toxic prompts and personally identifiable information leakage. (cloudflare.net) Cloudflare has also been telling investors that AI is becoming central to its business, not peripheral to it. In its February 10, 2026 earnings release, the company reported fourth-quarter revenue of $614.5 million, up 34% from a year earlier, and Chief Executive Matthew Prince said “AI and agents” were driving demand across Workers, security and networking products. (cloudflare.com) That helps explain why Anthropic’s announcement hit Cloudflare from two directions at once. It raised the prospect that frontier model companies could move deeper into security work, while also reminding investors that Cloudflare still depends on selling the infrastructure and guardrails around the AI boom. (anthropic.com) (cloudflare.com) Anthropic framed Project Glasswing as a defensive coalition, not a commercial replacement for security vendors. Several companies that sell cybersecurity products, including CrowdStrike and Palo Alto Networks, are participating in the program rather than opposing it. (anthropic.com) (cnbc.com) Anthropic also put real money behind the effort, saying it will provide up to $100 million in usage credits and $4 million in donations to open-source security groups. That scale made the announcement look less like a lab demo and more like an attempt to shape how AI-powered cyber defense is deployed. (anthropic.com) By Friday’s close, Cloudflare’s shares had fallen to $166.99 from $193.05 the prior close. The bigger question after this week is whether security companies end up as Anthropic’s customers, its partners, or its competitors in the next round of AI tooling. (stockanalysis.com)