Sources: CISA's Attack Surface Evaluation team is using Mythos to audit government code repositories and has already uncovered a large number of vulnerabilities
THE SO WHAT
If Anthropic’s Mythos is already surfacing large vulnerability volumes in federal code, AI code review just moved from “nice-to-have” to baseline security hygiene. Any org with a meaningful internal codebase should assume their current scanning stack is underpowered and budget for AI-augmented auditing this year.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIIllinois Drops the Hammer on AI Companies
State-level AI enforcement is moving from rhetoric to penalties. If you operate or sell into Illinois, assume AG and regulator scrutiny on data use and model behavior is now a real operational constraint, not a theoretical risk.
Applied AIUtah let an AI renew prescriptions without a doctor, and its medical board wants it stopped
Utah’s experiment with AI-only prescription renewals—and the medical board’s backlash—shows regulators will tolerate workflow augmentation long before full clinical substitution. If you’re building in health, design for physician-in-the-loop from day one or expect program shutdown risk.
Applied AIVercel CEO Guillermo Rauch on the fight to split off models from agents
Rauch’s point on price/performance in production is the real story—teams are starting to treat models as swappable commodities and agents as the durable asset. Architect your stack so you can change models without rewriting workflows, or you’ll overpay every time the pricing curve shifts.
Applied AIHigh AI Capex Demand a 'Multi-Year' Cycle, Says Rudina Seseri
If AI infra capex is a multi-year cycle, not a spike, then the cost of waiting goes up—capacity, pricing, and talent will be set by those committing now. Operators should map where they truly need proprietary compute versus where they can ride hyperscaler platforms and avoid getting dragged into an arms race they can’t win.