Anthropic leak and a cybersecurity sell-off

Market leaders in cybersecurity solutions, CrowdStrike, Zscaler, and Palo Alto, are posting several-percent losses in today’s session, with declines averaging around 6%.

This move is the aftermath of a leak concerning new products from Anthropic, the second-largest company in the AI sector by valuation and scale of operations.
Claude Code Security is a new addition to its developer toolset, intended to help engineers design safer systems and applications. This is something that, according to many industry representatives, has so far been a weak point of AI tools, including Claude Code. Many market participants believe that greater involvement of artificial intelligence in securing systems, networks, and software will reduce the market share of cybersecurity companies.

At the same time, a question arises: how effective will these companies’ efforts to protect users from cybercrime be if AI adapts to searching for vulnerabilities and gaps in security systems? Are the market’s concerns about these companies’ condition justified?

A different view is presented, among others, by analysts at the World Economic Forum. In their 2026 report, they argue quite clearly that AI increases, rather than decreases, corporate demand for cybersecurity services. The nature of threats is fundamentally changing: the focus is shifting toward agent-based solutions, identity verification, and automation. However, companies such as Palo Alto and Zscaler are pioneers in these areas, so there is no clear reason to price in worse outcomes for them.

There is also a second dimension to the ongoing sell-off, driven more by conjecture and speculation than by a sober assessment of the impact on business models. This dimension involves leaks about “Claude Mythos”, a new tool from Anthropic whose capabilities are unknown but are said to be “unprecedented.”

At this point, it is worth pausing to reflect on what is happening. Cybersecurity companies are being sold off because Anthropic has supposedly built tools so powerful that their business models are being called into question. And yet hardly anyone seems to be asking why we are learning about these tools through a data leak in the first place.

The conclusions suggest themselves: either the company cannot use its own tools effectively, or the tools are not as capable as Anthropic’s representatives claim, or the leak is intentional and meant to shape sentiment around AI, which in recent weeks appears to be losing the battle for capital and faith in the continued success of the “revolution.” None of these scenarios justifies sell-offs of this magnitude in an industry tied to cybersecurity.