G20 Leaders Want Answers On Claude Mythos, Anthropic Ready To Engage

Anthropic’s powerful new AI model, Claude Mythos, is no longer just a Silicon Valley talking point. The AI system has now caught the attention of some of the world’s most powerful financial regulators and governments, with concerns growing that advanced AI could expose dangerous weaknesses inside global banking and payment systems. According to a report by the Financial Times, Anthropic has agreed to brief finance ministries, regulators and central banks linked to the Financial Stability Board (FSB) about cybersecurity vulnerabilities identified by its Claude Mythos Preview model. The move comes as governments worldwide race to understand whether next-generation AI systems could accidentally become tools for large-scale cyberattacks.Why Global

Regulators Are ConcernedThe discussions reportedly began after a request from Andrew Bailey, chair of the Financial Stability Board. The FSB includes central bankers, regulators and finance ministry officials from major G20 economies including India, the US, the UK, China, Japan and Germany.Officials are reportedly worried that Claude Mythos could detect serious flaws inside banking infrastructure, payment networks and financial software faster than humans can respond to them.The bigger fear is not just the discovery of vulnerabilities, but what happens if malicious actors gain access to such insights before banks or institutions patch the problems.That concern has already pushed regulators to encourage financial institutions to improve cybersecurity defences and accelerate software updates.The Financial Stability Board is also reportedly preparing a separate report outlining “sound practices” for AI adoption in the financial system. According to the Financial Times report, the consultation paper is expected next month.India Has Already Started PreparingIndia, too, has started taking the issue seriously. Back in April, Finance Minister Nirmala Sitharaman reportedly held discussions with commercial bank executives, Reserve Bank of India officials and representatives from CERT-In to review emerging cyber threats linked to advanced AI systems.The meeting reportedly focused on building stronger real-time threat intelligence sharing systems to handle AI-driven cybersecurity risks more effectively.The discussions reflected growing fears that AI models like Mythos could dramatically speed up how cyber threats spread across financial networks.Humanoid Robots Aren’t Replacing Humans Yet: Intern Wins This ChallengeAccess To Mythos Remains RestrictedAnthropic has been extremely cautious about how widely Claude Mythos is distributed. The OpenAI rival currently offers access to the model only through its tightly controlled “Project Glasswing” programme. Reports suggest only around 40 organisations currently have access, most of them based in the United States. Those reportedly include major firms such as Amazon, Microsoft and JPMorgan Chase.The limited rollout is intended to help organisations safely identify vulnerabilities and deploy fixes before broader access becomes available. According to reports, Anthropic also agreed not to expand public access to Mythos after concerns raised by the White House.