Anthropic Mythos model can find and exploit 0-days

72.4%: Anthropic's Mythos Preview produced working exploits in 72.4% of tests and reportedly identified 'thousands' of high- and critical-severity zero-day vulnerabilities across major operating systems and web browsers. Anthropic withheld a public release and provided a limited preview to industry partners under Project Glasswing (participants include AWS, Apple, Google, Microsoft, NVIDIA, CrowdStrike, Palo Alto Networks, Broadcom, Cisco, JPMorganChase, Linux Foundation), subsidizing access with up to $100M in usage credits and $4M in donations to open-source security groups. Investment implications: elevated systemic cyber risk that could drive increased spending (positive for security vendors and cloud providers) while imposing patching/operational costs and potential regulatory scrutiny for affected OS/browser vendors.

Analysis

Treat this as an acceleration of a recurring budget cycle rather than a one-off shock: enterprises will move more spend from perimeter controls to managed detection, rapid patching pipelines, and bug-bounty/ADP-style programs. Expect commercial providers that can turn rapid dynamic analysis into callable API products to win the bulk of incremental spend; that market reallocation can add mid-to-high single-digit revenue growth to best-in-class vendors within 12–24 months without a comparable increase in GAAP margins in year one.

There is a distinct supply-chain bifurcation forming: platform/cloud providers that can bundle ‘hardened’ services gain a sticky revenue premium but also concentration of liability and remediation cost. This will push large enterprises to multi-cloud redundancy for critical workloads within 6–18 months, increasing inter-cloud data egress and security orchestration demand — a win for vendors that monetize cross-cloud telemetry and orchestration.

Hardware demand will reorient toward specialized accelerators and high-memory GPUs for large-scale fuzzing and formal verification workloads; that structural demand is multi-year and likely to favor incumbents with software ecosystems and tooling partnerships. Conversely, legacy silicon and slow-moving on-prem vendors face compressive pricing pressure if they cannot pair hardware SKU growth with differentiated security software or managed services.

Policy and insurance are an underappreciated catalyst: expect insurers to tighten cyber policy terms and raise premiums over 6–12 months, which will force C-suite prioritization and capital allocation to security — a steady secular tailwind for repeat-revenue security vendors but a potential fiscal headwind for cloud providers carrying remediation liabilities.

Key Decisions for Investors

• Long CRWD (CrowdStrike) — buy 9–15 month calls or accumulate stock over 3–6 weeks. Rationale: highest leverage to recurring endpoint + cloud telemetry monetization; target >25% upside if ARR acceleration re-rates multiple. Risk: execution/competitive displacement; set stop at -18% from entry.
• Long PANW (Palo Alto) with protective put — buy shares and hedge with 6–9 month put ~10–12% OTM. Rationale: franchises a broad portfolio (NGFW, cloud security, orchestration) that benefits from higher enterprise spend; asymmetric payoff if adoption accelerates. Hedge preserves capital vs tactical reversal.
• Long NVDA calls (6–12 month) sized as a satellite position. Rationale: durable multi-year uplift in accelerator demand for large-scale security research and verification workloads can boost data-center TAM and pricing power. Risk: GPU cycle volatility; cap position size to <3% NAV and use spreads to fund premium.
• Short INTC or buy INTC 3–9 month put spread as a hedge against slower execution on accelerator roadmap. Rationale: secular pressure on legacy server CPU share and weaker competitiveness in accelerators creates limited near-term upside. Risk: product-cycle beat or government procurement could reverse; limit position to <2% NAV.