Financial services organisations and watchdogs for the sector must embrace AI to counter the threat of AI attacks, a leading Swiss financial regulator has told Reuters.

Marlene Amstad, president of Swiss market regulator FINMA, spoke to the publication on the sidelines of a crypto-market hackathon to advocate the use of AI among financial sector organisations.

"As hackers move faster, banks must adapt by patching vulnerabilities more rapidly," Amstad told Reuters.

Although AI-powered attacks have been more of a theoretical concern to date, cybersecurity experts warn that they will rise in prominence by 2027. On 24 June, cybersecurity and intelligence agencies from across the Five Eyes intelligence alliance warned that sophisticated AI attacks will come in months, not years.

Access to models such as Anthropic’s Mythos has helped to expose vulnerabilities and contextualise the threat posed by AI.

Amstad added that in the future, these models must also be used to assess and strengthen IT systems in banking before they are deployed and that to achieve this, regulators would need to continue to use the most powerful AI tools on the market.

"Switzerland must retain access ⁠to the most advanced AI models,” she said.

In recent months, releases of advanced AI models have sparked fears that AI could soon be used for more advanced cyber attacks.

In April, the European Central Bank held a meeting with banks to discuss the potential risks posed by Anthropic’s Mythos, which can be used to detect previously undisclosed cybersecurity vulnerabilities.

Anthropic has since disabled both Mythos and its general-purpose counterpart Fable, following a US government order to restrict access to US citizens.

The US government has likewise imposed a staggered release for OpenAI’s newest cyber model GPT-5.6 Sol, with initial access only available to hand-picked “trusted partners”.

---