The fear is not that an AI will suddenly “take over Wall Street.” It is subtler and, for that reason, more plausible. On June 10, 2026, the Financial Stability Board (FSB) opened a consultation on responsible AI adoption in finance, after its June 1 plenary had already highlighted new vulnerabilities to financial stability and said a final report would be delivered to G20 finance ministers and central bank governors in October. Comments are due by July 22, 2026. That timetable alone signals urgency: supervisors no longer see agentic AI as a futuristic curiosity, but as a live policy problem. (fsb.org)
The reason is straightforward. The FSB defines agentic AI as systems able to autonomously perform complex and extended tasks. According to the Cambridge Centre for Alternative Finance, 52% of surveyed financial firms are already piloting or deploying agentic AI, with 23% at scaling or transformation stages and fintechs ahead of traditional institutions. Regulators, however, report much lower agentic adoption—28% overall—suggesting a widening supervisory gap. In finance, that gap matters because software does not merely generate text; it can shape fraud detection, customer interaction, operational processes, and decision support at machine speed. (fsb.org)
Why the alarm? The FSB warns that highly autonomous agents can take unauthorised or erroneous actions, expose sensitive data, and disrupt connected systems, and that these risks can materialise at great speed. Its 2024 report had already identified third-party concentration, herd-like market correlations, cyber risk, and weak model governance as vulnerabilities with systemic potential. Cambridge’s 2026 survey reinforces that concern: OpenAI, Google, and Anthropic were the most prominent foundation-model providers among respondents, raising the spectre of common dependencies across the sector. If too many firms rely on similar models, data sources, and cloud infrastructure, one failure—or one shared blind spot—can propagate unusually far, unusually fast. (fsb.org)
That is why regulators are moving now. The FSB is pushing firms to define clear boundaries for AI agents, keep humans in command for highly autonomous systems, maintain kill switches and audit trails, and manage concentration risk in third-party AI providers. It even suggests adapting HR-style controls so that agents are treated, in effect, like “synthetic employees.” The debate, then, is no longer whether agentic AI will enter finance; it already has. The real contest is whether governance can evolve quickly enough to match the speed, scale, and opacity of the machines now entering the system. (fsb.org)










