Cybersecurity Stocks Drop: Anthropic Tests Powerful AI Model

Cybersecurity stocks declined sharply on Friday following revelations that Anthropic has begun testing “Mythos,” an extraordinarily powerful new AI model with advanced vulnerability-discovery capabilities.

Anthropic is actively trialing a new tier of artificial intelligence models codenamed “Capybara,” with the flagship model operating under the moniker “Mythos”.

Internal documents indicate that Mythos significantly outperforms the company’s previously most advanced model, Claude Opus 4.6, in academic reasoning, software coding, and cybersecurity benchmarks.

Characterized by Anthropic as a “step change” in performance, the model is currently restricted to a highly vetted group of early-access customers.

The company has deliberately paced its release due to the unprecedented nature of the model’s technical proficiencies and its capacity to surface complex code flaws.

Market Impact and Equity Declines

The financial markets reacted swiftly to the news, driven by renewed anxieties that advanced AI software tools will aggressively compete with industry incumbents.

The Global X Cybersecurity ETF plummeted by 4.5% on Friday, establishing its lowest closing position since November 2023 and pushing its year-to-date decline beyond 21%. This sell-off reflects a growing market sentiment that autonomous AI agents may disrupt traditional enterprise security architectures.

The primary catalyst for the industry disruption is the dual-use nature of the Mythos model. During testing, Mythos demonstrated the ability to autonomously surface previously unknown vulnerabilities, potentially zero-days, within live production codebases.

Anthropic’s internal assessments explicitly warn that Mythos is “currently far ahead of any other AI model in cyber capabilities”. The company cautioned that such technology presages an upcoming wave of offensive AI tools capable of exploiting vulnerabilities at a pace that far exceeds the patching efforts of human defenders.

This has raised alarms regarding the potential for large-scale, automated cyberattacks if the model’s safeguards are bypassed or replicated by adversaries.

This latest development compounds existing market friction introduced in February 2026, when Anthropic launched Claude Code Security. That tool shifted the paradigm from static, rule-based pattern recognition to AI-driven dynamic reasoning, analyzing codebases much like a human security researcher to trace data flows and identify complex flaws.

The prospect of highly resourced threat actors leveraging models like Mythos to weaponize vulnerability discovery remains a documented, critical risk. Anthropic previously disclosed that a state-sponsored entity from China had already attempted to employ earlier versions of Claude to automate attack sequences.

As AI rapidly transitions from a supportive utility to an autonomous vulnerability hunter, traditional cybersecurity vendors face immense pressure to overhaul their detection engines or risk obsolescence against machine-speed threats.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.