AI Safety Talks: Banks and Governments Get the Inside Story

Artificial‑intelligence companies are stepping up to explain how their new tools could threaten global finance. Anthropic, the maker of Claude Mythos Preview, has agreed to meet with top finance ministries and central banks. The meetings aim to warn about cyber weaknesses that the AI model may reveal.

• Bank of England chief Andrew Bailey, who runs the Financial Stability Board (FSB), requested the briefing.
• The FSB is a G20 group that monitors worldwide financial health and will publish next month a guide on safe AI use, underscoring the board’s urgency.

Only a few people can see Claude Mythos right now. This limited launch worries regulators outside the United States, who fear that different rules could let bad actors exploit gaps. Anthropic said it will give high‑level talks to European and other non‑U.S. bodies, including the European Commission.

The International Monetary Fund warns that advanced AI can accelerate attacks. If a hacker finds a flaw, the model can spread it faster than companies can fix it, giving attackers an edge.

Other AI firms are also racing to create cyber tools. OpenAI launched GPT‑5.5-Cyber for security teams, adding pressure on competitors to improve safety. Meanwhile, the UK government is courting Anthropic for a London office and a possible stock listing.

Anthropic plans to go public in 2026. The company’s moves suggest that AI safety is becoming a global priority, not just a technical issue.