Top AI Product

So Anthropic just made a pretty big move. Yesterday they officially launched [Claude Code Security](https://www.anthropic.com/news/claude-code-security), their first dedicated cybersecurity product, and honestly, the early results are kind of wild.

Here’s what caught my attention. This thing isn’t another static analysis tool that matches patterns against a list of known bad code. It’s built on Opus 4.6 and actually *reads* code the way a human security researcher would. It traces data flows across components, understands how different parts of a system interact, and goes after the stuff that rule-based scanners flat out miss — logic flaws, broken access controls, the kind of subtle bugs that only show up when you really understand what the code is supposed to do.

The numbers are hard to ignore. Anthropic’s [Frontier Red Team](https://red.anthropic.com/2026/zero-days/) tested Opus 4.6 against some of the most heavily audited open-source codebases out there — projects that have had fuzzers hammering them for years. It still found over 500 previously unknown high-severity vulnerabilities, including buffer overflows in OpenSC and CGIF, and a particularly sneaky Ghostscript bug that Claude only caught by digging through git commit history after traditional tools came up empty. Some of these bugs had been hiding in plain sight for decades.

What I find genuinely interesting is the self-verification step. Every finding goes through a multi-stage process where Claude basically tries to prove itself wrong before reporting anything. That’s a smart way to deal with the false positive problem that makes most automated security tools so exhausting to use in practice.

The launch got a [Fortune exclusive](https://fortune.com/2026/02/20/exclusive-anthropic-rolls-out-ai-tool-that-can-hunt-software-bugs-on-its-own-including-the-most-dangerous-ones-humans-miss/) and quickly spread across [Hacker News](https://news.ycombinator.com/item?id=44814264), [DevOps.com](https://devops.com/anthropic-adds-automated-security-reviews-to-claude-code/), and Bloomberg — where the headline was about [cyber stocks sliding](https://www.bloomberg.com/news/articles/2026-02-20/cyber-stocks-slide-as-anthropic-unveils-claude-code-security) on the news. That tells you something about how seriously the market is taking this.

Right now it’s a limited research preview for Enterprise and Team customers. But there’s one detail I really appreciate: open-source maintainers get free expedited access. Given how many critical projects run on volunteer effort and zero security budget, that feels like exactly the right call.