AI companies build powerful tech — but they are expert marketers

Anthropic’s announcement of Claude Mythos and the Project Glasswing consortium set off immediate attention in markets and the cybersecurity community, with the firm saying the model exposed thousands of vulnerabilities but will not be widely released.

According to Anthropic, Mythos Preview uncovered long-standing flaws — including a 27‑year‑old issue in OpenBSD — and the company invited a group of major cloud, chip and security firms to review and remediate findings under Project Glasswing. Participating organizations include Amazon Web Services, Apple, Broadcom, Google, Microsoft, Nvidia, CrowdStrike, Palo Alto Networks and JPMorgan Chase.

The market reaction was tangible: software and cybersecurity equities experienced selling pressure as investors re-evaluated competitive moats and risk exposure. Industry reports and market commentators noted declines in software-heavy indices and selected security vendors, reflecting a repricing of how AI-driven tooling may disrupt incumbent security business models.

In the broader context, the episode highlights two trends relevant to global investors: the dual‑use nature of frontier AI that can both find and weaponize software flaws, and the commercial strategy of using scarcity and controlled releases to amplify demand and shape narratives. Both dynamics complicate regulatory oversight and raise questions about disclosure norms for dual‑use capabilities.

Analysts advise a measured approach: monitor verified technical assessments, partnership disclosures and cloud providers’ adoption terms, while watching for sector rotation between defensive cyber vendors and beneficiaries such as major cloud platforms and chipmakers. For portfolio managers, the immediate task is sizing exposure and distinguishing marketing rhetoric from demonstrable, auditable capability.