Palo Alto Networks says Mythos found 24+ critical bugs using $1M+ in tokens; Anthropic subsidizes Mythos but some companies plan to boost their Mythos budgets
When Palo Alto Networks earlier this year began testing Anthropic's Claude Mythos to comb through its own source code, it didn't take long to see the future of cybersecurity.
Context & Ripple Effects
Anthropic positioned Mythos Preview as a restricted tool for organizations maintaining critical software after reporting that it had found high-severity vulnerabilities across widely used software. Subsequent coverage says the model can turn disclosed vulnerabilities into working exploits much faster, while CISA is using it to audit government repositories.
Palo Alto Networks' reported results add a concrete enterprise-codebase use case to that arc: Mythos is being used not only to demonstrate vulnerability discovery, but to find critical flaws in a customer's own code despite substantial inference costs.
First-order effects
- Palo Alto Networks gains a larger queue of critical bugs to validate and remediate, while Anthropic gets a prominent production-style reference point for Mythos's security-auditing value.
- The reported million-plus-token usage makes model cost an immediate procurement issue; Anthropic's subsidies reduce the near-term barrier, but companies planning larger budgets signal that sustained use will require dedicated spend.
Second-order effects
- Security teams and software owners evaluating AI-assisted code review will face pressure to test comparable tools, particularly for high-value or critical code, while budgeting shifts from occasional assessments toward ongoing model-driven audits.
- A faster vulnerability-finding workflow can expand downstream demand for human triage, patching, and remediation capacity; finding more issues does not itself resolve the operational backlog.
Third-order effects
- If restricted access broadens and reported results prove repeatable, vulnerability discovery may become a continuous, model-assisted capability rather than a periodic specialist exercise—raising both software assurance expectations and the cost of keeping codebases current.
- The same capability creates a dual-use governance challenge: coverage already indicates Mythos can accelerate exploit development from public disclosures, so access controls and deployment policies may become as consequential as detection performance.
The trend: Mythos is one data point in the shift from AI as a coding assistant to AI as a high-cost, tightly governed security-analysis system that can accelerate both defense and offensive exploitation.