Anthropic faces mounting pressure over access to its powerful Mythos artificial intelligence model, with major banks like JPMorgan Chase, Bank of America, and Goldman Sachs granted early access while smaller and international lenders were excluded. This selective distribution has raised concerns about fairness and cybersecurity preparedness across the global financial system.

The controversy centers on Project Glasswing, Anthropic's limited-access program for studying Mythos's dangers. Initially, participants worried about confidentiality restrictions that prevented sharing cybersecurity findings with non-participants. However, the company clarified that members can now disclose discoveries, best practices, and code to other companies, industry bodies, regulators, and media. This shift aims to maximize defensive impact against potential AI-powered threats.

Regulatory warnings have intensified the debate. Bank of England Governor Andrew Bailey recently warned that Mythos could 'crack the whole cyber-risk world open,' while European Central Bank official Frank Elderson cautioned banks to strengthen defenses against AI-enabled breaches. Industry groups are coordinating responses through the Financial Service Sector Coordinating Council, with Anthropic joining the Analysis and Resilience Center for Systemic Risk.

Competitors are moving quickly to capitalize on the moment. OpenAI recently launched GPT-5.5-Cyber with wider distribution plans, and French startup Mistral is pitching its own AI model to European banks. This growing ecosystem suggests that AI-powered cybersecurity tools will become increasingly central to financial sector risk management.