"AI panic" emerges, causing a collective plunge in US stock cybersecurity concepts

On February 20th, local time, the three major U.S. stock indices all opened lower, then surged sharply during the session to turn positive. As of the latest update, the Dow Jones Industrial Average rose 0.23%, the Nasdaq increased 0.93%, and the S&P 500 gained 0.54%.

Large-cap technology stocks mostly rose, with Google up 3.5%. In news, on February 19th local time, Google released its latest foundational large model Gemini 3.1 Pro. Compared to the Gemini 3 Pro released in November last year, the new model’s inference performance has doubled.

The Nasdaq China Golden Dragon Index opened lower but recovered to turn positive, currently up 0.06%.

Meanwhile, cybersecurity concept stocks experienced a collective sharp decline. By the close, SailPoint dropped 9.44%, Okta fell 9.18%, Cloudflare declined 8.05%, CrowdStrike decreased 7.95%, and Zscaler fell 5.47%.

The Global X Cybersecurity ETF dropped nearly 5%, marking its lowest closing level since November 2023.

Earlier, Anthropic PBC announced Claude Code Security, a new AI tool capable of scanning codebase vulnerabilities and suggesting fixes. Claude Code Security not only scans for known patterns like traditional static analysis tools but also reads code like a human security researcher, tracking data flows, understanding component interactions, and flagging subtle logical flaws often overlooked by rule-based scanners. Each finding undergoes a multi-stage verification process before reaching human analysts, and nothing is patched without developer approval.

What concerns investors is that Anthropic PBC stated its Claude Opus 4.6 model discovered over 500 vulnerabilities in internal testing within open-source code repositories—vulnerabilities that have survived decades of expert review. The company also mentioned in its blog that it uses Claude to review its own code and finds it extremely effective in protecting Anthropic systems.

Anthropic PBC indicated that the new tool is currently available only as a limited research preview for enterprise and team customers, with open-source maintainers granted priority access. It focuses on code auditing and vulnerability detection—not real-time endpoint protection, identity management, or zero-trust networking, which remain core business areas for companies under heavy attack.

Dennis Dick, trading director at Triple D Trading, noted that software stocks have been steadily sold off, and on that day, cybersecurity stocks experienced a mini flash crash triggered by a headline.

The ability of Anthropic PBC to detect vulnerabilities that have gone unnoticed for decades may undermine the long-term value of existing enterprise security suites. Market participants are concerned that AI-native tools are shifting security focus from passive monitoring to proactive automated remediation.

Source: Public information and market data