
CrowdStrike
US cybersecurity firm; one of twelve Glasswing partners granted Mythos AI access.
Last refreshed: 10 April 2026 · Appears in 2 active topics
Why did Anthropic give CrowdStrike access to its most powerful AI model?
Timeline for CrowdStrike
Mentioned in: MIT economist: AI layoffs are a cover story
AI: Jobs, Power & MoneyCut 500 roles with CEO attributing cuts directly to AI flattening the hiring curve
AI: Jobs, Power & Money: Dell cuts 11,000 on record revenueMentioned in: Balderton leads Geordie's $30m AI round
UK Startups and InnovationMentioned in: Nvidia adds 6,000 staff as rivals cut
AI: Jobs, Power & MoneyMentioned in: UNC6780 takes Cisco AI Defense source code
Cybersecurity: Threats and DefencesWhy did Anthropic give CrowdStrike access to Claude Mythos?
Is CrowdStrike using AI to find zero-day vulnerabilities?
What happened with CrowdStrike in 2024?
Background
CrowdStrike is one of twelve organisations selected by Anthropic for Project Glasswing, the controlled-access programme through which Claude Mythos Preview was released on 8 April 2026. Partners received $100 million in model usage credits and exclusive access to a system that scored 83.1% on the CyberGym vulnerability reproduction benchmark.
CrowdStrike is a leading US cybersecurity firm best known for its Falcon endpoint protection platform, which uses AI and behavioural analysis to detect and respond to threats across cloud, endpoint, and identity environments. The company gained global recognition , and notoriety , in July 2024 when a faulty Falcon sensor update caused an estimated 8.5 million Windows devices to crash, triggering one of the largest IT outages in history.
CrowdStrike's inclusion in Glasswing reflects both its technical credibility in AI-assisted threat detection and the high-stakes Nature of giving offensive-grade AI to a firm that already operates deep inside enterprise security stacks worldwide. Critics noted that access to Mythos by a firm with CrowdStrike's reach amplifies both the defensive potential and the systemic risk if the model or its access is misused.