The Canadian Bankers Association has responded to the Government of Canada’s consultation on AI transparency, arguing that existing banking, privacy and supervisory frameworks should remain the primary means of managing AI risks in the financial sector. It sees no demonstrated need for standalone AI transparency rules for banks. Any additional measures should address a material regulatory gap, operate through existing regulators and apply only where AI could materially affect or reasonably mislead individuals. Disclosure should focus on substantially AI generated or materially modified content that could be mistaken for an authentic person, event or communication, rather than incidental AI use. Customers should also be informed of direct AI interactions when the technology is not apparent and could affect their expectations or decisions, but disclosure should not compromise fraud prevention, anti money laundering monitoring, cybersecurity controls or lawful investigations. Obligations should reflect each party’s knowledge and control, with providers supplying technical and provenance information and deployers determining disclosures for specific uses. Any AI incident reporting should use existing supervisory channels, apply a high materiality threshold and avoid duplicate reports to multiple authorities. For AI agents, banks should be able to verify whom an agent represents, its permitted activities and whether its authorization remains valid, but prescriptive rules on conduct or liability would be premature. The association also supports interoperable supplier documentation and industry standards for third party systems while opposing requirements to disclose proprietary models, architecture or training data.

Original source