In its latest report, the Financial Stability Board (FSB) urges financial institutions to implement protective measures to manage risks posed by artificial intelligence, particularly systems based on so-called AI agents. The FSB warns that the growing autonomy of AI systems could amplify risks within the financial sector as adoption has accelerated.
What are AI agents and where are they used?
AI agent–based solutions are systems that can plan, reason and execute tasks with limited human oversight. According to the FSB, such solutions are already in use across the financial sector for activities including fraud detection, customer service, and back‑office operations.
A survey by the Cambridge Centre for Alternative Finance found that 52 percent of respondents in the financial sector reported active use of agent‑based AI. Of those, 23 percent deploy the technology at an organizational level, while 29 percent are running pilot tests.
Risks highlighted by the FSB
The FSB states that autonomous AI can give rise to risks that can become realities very quickly. The report lists specific concerns such as:
- unauthorized or unlawful actions;
- data leakage; and
- operational disruptions in interconnected systems.
The report emphasizes that intelligent agents pose particular challenges for human oversight, since they may take actions that diverge from an organization’s intent and that staff might not detect or be able to intervene against in time.
The FSB also notes heightened regulatory concern following the market release of Anthropic’s Mythos system, which has been assessed by experts as presenting significant cybersecurity challenges for the banking sector.
FSB recommendations and consultation deadline
The FSB’s guidance is non‑binding and is open for public comment until July 22. Key recommendations include:
- Financial institutions should set clear boundaries for AI use and document which tasks AI agents perform.
- Robust safeguards should be built into systems to mitigate unauthorized operations, data leakage, and systemic effects.
- Human approval should be mandatory for high‑risk operations, for example financial transactions above certain thresholds.
- The report proposes that firms consider treating AI agents as "synthetic employees," extending human‑resources‑style checks and related processes to these systems to help mitigate risks.
Why this matters for the financial sector
The FSB’s findings matter because the interconnected nature of financial systems and the high volumes involved mean errors or malicious events can quickly have systemwide consequences. The autonomous behaviour of AI agents and the difficulties of monitoring them make their broad deployment in banking and financial services particularly risky.
While the FSB’s recommendations are not legally binding, they provide guidance for regulators and market participants on safer deployment and operation of AI technologies. The consultation on the report and its recommendations may prompt further debate about the appropriate level of human control and the types of internal checks required to integrate AI agents safely into the financial sector.
