The Financial Stability Board pulled back the curtain on one of its highest-profile artificial intelligence workstreams this month, posting the full batch of comments it received on a global consultation about how banks, insurers and other financial institutions should approach AI adoption. For an industry still working out where automated underwriting, claims triage and algorithmic pricing sit on the spectrum between innovation and liability, the release offers a rare, unfiltered look at how the sector itself is framing the debate.
FSB opens the vault on AI adoption feedback
On 10 June 2026, the FSB published Sound Practices for Responsible Adoption of Artificial Intelligence (AI): Consultation report, kicking off a formal comment period that stretched across the summer. Interested parties were invited to provide written comments by 22 July 2026, giving regulators, trade associations, insurers and individual firms roughly six weeks to weigh in on the draft framework before the window closed. Rather than summarising the input internally, the board opted for full disclosure: The public comments received are available below on its site, posted alongside the original text so readers can measure each submission against the passage it responds to. That level of openness is unusual for a standard-setting body whose deliberations more often stay behind closed doors until a final text is ready, and it gives trade associations, insurers and analysts an early read on where consensus is forming and where it is not.
The gesture doubled as a thank-you note to an industry that does not always get visibility into how its feedback lands with a standard-setter. The FSB thanks those who took the time and effort to express their views, the board wrote, closing out a process that drew responses from across the financial sector rather than from banking alone, and that now sits in the public record for competitors, supervisors and vendors alike to pore over.
Twelve practices anchor the emerging playbook
At the centre of the consultation is a structured governance framework rather than a single rulebook. the report proposes a menu of 12 sound practices that financial institutions could apply in their organisation-wide AI governance and management, spanning everything from model risk oversight to vendor due diligence and incident reporting. The framework’s authors were explicit that firms need more than a static checklist to keep pace with the technology: financial institutions need to understand and remain updated on the opportunities and risks of AI, and respond with the appropriate adoption strategy and guardrails to manage evolving associated risks. That living-document instinct echoes what prudential supervisors have been signalling elsewhere in the industry, including enhanced market surveillance tools for insurers rolled out by Germany’s regulator and the ongoing oversight discussions among state regulators that have become a fixture of US insurance supervision.
Insurers weigh AI’s upside against amplified risk
For insurance executives, the consultation lands squarely inside a debate that has already spilled into supervisory colleges from Basel to Canberra. The FSB’s own framing of the stakes is unambiguous: Financial institutions are leveraging AI to transform operations and services, but its rapid adoption may also amplify or introduce risks that need to be identified and managed appropriately. Prudential regulators have been moving in parallel on that exact tension — Australia’s supervisor has pressed for a push for stronger board-level AI oversight across insurers, a stance that lines up closely with the FSB’s systemic framing: At the financial system level, responsible AI adoption reduces risks to financial stability.
None of this is meant to chill adoption outright. Handled well, the FSB argues, the payoff is real: Responsible AI adoption allows financial institutions to harness opportunities and benefits while minimising associated risks. That balancing act is precisely why the sound-practices menu was pitched as a shared reference point rather than a prescriptive rulebook — the board says The FSB has developed sound practices to help all types of financial institutions navigate benefits and risks responsibly as they adopt AI, language broad enough to cover mutual insurers and reinsurers alongside deposit-takers and asset managers. For carriers already navigating overlapping national regimes, a common international reference point carries obvious appeal, even if it stops short of binding rules. Underwriting algorithms, automated claims triage and behaviour-based pricing all sit within scope of that broader framing, and each raises its own version of the same question the FSB is wrestling with at system level: how much autonomy should a model be given before a human needs to sign off. Boards that have already built model-risk committees for actuarial and capital models are, in many cases, simply extending that muscle memory to cover generative and predictive AI tools, rather than starting governance from scratch.
Final report due as debate over rules intensifies
The consultation draft was not built in isolation. The FSB says the underlying report incorporates insights from a range of stakeholders across the financial system, including financial institutions and their technology vendors, a methodology that mirrors the multi-stakeholder approach already visible in the debate over state versus federal rules playing out among US insurance regulators, and detailed further in the original consultation report on responsible AI adoption.
With the comment window closed and submissions published, attention now turns to what the FSB does with the feedback. On timing, the board was direct, confirming in its summary of next steps for the sound practices work that The FSB expects to publish the final report in the coming mopnths. Insurers, brokers and their technology vendors now have a narrow window to see whether the concerns they raised — on liability allocation, model explainability or vendor accountability — made it into the version that ships next, and whether a globally coordinated standard will ease or complicate compliance against the patchwork of national rules already taking shape. For multinational groups writing business across several jurisdictions at once, the appeal of a single reference framework is straightforward: fewer diverging definitions of what counts as adequate AI governance, and one fewer gap to bridge when a group-wide model gets deployed locally by a subsidiary. Whether the finished text delivers that consistency, or simply adds another layer alongside national supervisors’ own expectations, will not be clear until the final version lands.
