Global regulators expand AI oversight for financial risk controls

Global regulators expand AI oversight for financial risk controls
AI oversight expands globally

As concerns grow over artificial intelligence's potential to destabilize markets and critical infrastructure, financial supervisors are increasingly exploring how to contain those risks with their own digital tools. The debate extends beyond possible asset bubbles and cyber threats to whether regulators can move faster than lawmakers by deploying agentic AI for monitoring and enforcement.

Highlights

  • Global regulators are intensifying efforts to manage AI-related financial risks, focusing on cyber threats and cloud concentration without waiting for new legal frameworks.
  • Supervisory bodies from over a dozen countries embraced creative use of existing powers at the Cambridge Agentic Regulator hackathon, reflecting urgency as AI outpaces legislative action.
  • Regulators globally, including in India, Kenya and the UAE, are developing their own AI tools and shifting focus from model explainability to outcome-based oversight, legal liability, and operational controls.

Supervisory tools and policy gaps

As reported by the Financial Times, regulatory officials and policy figures are intensifying discussions on how to address AI-related financial risks, including cyber threats to market infrastructure and concentrated exposure to cloud providers. The article says these debates are unfolding alongside broader international concern over AI governance, with some officials concluding that supervisors cannot wait for politicians to build new legal frameworks before acting.

That approach is already taking shape in practice. At a digital regulation meeting in Cambridge earlier this month, co-convened by former Bank for International Settlements head Agustín Carstens and Financial Conduct Authority chair Ashley Alder, participants took part in what was described as the first global Agentic Regulator hackathon. Dozens of regulators from financial and non-financial fields joined, and nearly 300 teams submitted projects.

One conclusion from those discussions is that existing supervisory powers may need to be used more creatively because AI is evolving faster than legislatures or diplomatic agreements. The article also points to recent moves in the U.S. and UK, including Treasury Secretary Scott Bessent's discussions with bankers about cyber vulnerabilities linked to Anthropic's Claude Mythos model and the Bank of England's pledge to supervise cloud companies as critical third parties.

Explainability debate and industry implications

Another emerging shift is in how regulators think about explainability, the principle that digital models should be transparent, interpretable and accountable. While that standard has long been central in finance, the article argues that enforcing it in AI may be difficult, prompting some supervisors to focus more on outcomes, legal liability and operational guardrails such as kill switches.

Regulators are also building their own AI systems to understand the technology and strengthen market oversight. Examples highlighted include tools designed to track capital flows, monitor herding in markets, detect fraud risks, identify vulnerabilities and oversee robo-advisers.

The experimentation is not limited to Western institutions. Some of the more inventive ideas are said to be coming from countries including India, Pakistan, Rwanda, Kenya and the United Arab Emirates, underscoring how AI supervision is becoming a broader competitive and policy issue for governments as well as for private sector technology groups.

In our earlier article on Anthropic’s seven-year cloud services deal with Akamai, we explained how the AI developer is securing long-term computing capacity as AI workloads surge. We also noted the contract’s scale (starting at $11.6 billion with an option to expand) and the equity warrant component, highlighting how demand for AI infrastructure is reshaping the cloud market beyond the largest hyperscalers.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.