Sources: Goldman Sachs, Citigroup, and Other Banks Are Testing Anthropic's Mythos Model Internally; JPMorgan Chase Is the only Bank Named in Project Glasswing (Bloomberg)

Why It Matters

Banks adopting powerful AI for cyber defense face heightened regulatory oversight, making AI security a strategic priority for the financial sector.

Key Takeaways

• Goldman Sachs and Citi test Anthropic's Mythos for cyber risk
• JPMorgan only bank named in Project Glasswing
• Treasury warns banks of Mythos cybersecurity threats
• Mythos deemed most capable model for cyber‑range tests
• CoreWeave secures multi‑year GPU cloud contract for Claude

Pulse Analysis

Anthropic’s Mythos model is quickly becoming a benchmark in AI‑driven cybersecurity. The model’s ability to simulate sophisticated attack vectors and complete end‑to‑end cyber‑range exercises has impressed banks seeking to harden their digital defenses. By integrating Mythos into internal threat‑hunting platforms, institutions like Goldman Sachs and Citigroup aim to detect vulnerabilities faster than traditional tools, positioning AI as a frontline asset in the fight against increasingly complex cyber threats.

Regulators are responding with unprecedented urgency. Treasury Secretary Scott Bessent and Fed Chair Jerome Powell summoned senior bank executives for an emergency briefing, warning that models such as Mythos could inadvertently create new attack surfaces if misused. Project Glasswing, the Treasury‑led initiative, highlights the need for clear guardrails and oversight as financial firms experiment with generative AI. JPMorgan Chase’s inclusion as the sole named participant signals both its leadership role and the heightened scrutiny it faces from policymakers.

The broader market is adapting to this AI security wave. CoreWeave’s multi‑year agreement to supply GPU capacity for Anthropic’s Claude models demonstrates growing demand for specialized compute infrastructure that can support both generative AI workloads and security‑focused applications. As banks scale AI deployments, partnerships with cloud providers will be critical to ensure performance, compliance, and resilience. The convergence of advanced AI models, regulatory pressure, and robust compute resources suggests that AI‑enhanced cyber defense will become a core component of financial institutions’ risk‑management strategies moving forward.