At a panel during the AI & Digital Bridge 2026 international forum, First Deputy Chairman Timur Abilkassymov outlined the Agency for Regulation and Development of the Financial Market of the Republic of Kazakhstan’s approach to supervising AI agents in finance. Oversight should cover the full process from objectives and instructions to decisions and outcomes, with priorities including algorithmic transparency, human control and protection against discrimination, hidden fees and forced add-on services. The approach builds on the agency’s broader work to digitize supervision and develop an AI regulatory framework. Abilkassymov distinguished manageable systems, in which people set trust levels, limits and conditions, from emerging autonomous agents capable of making decisions and assembling packages of products and services. He called for safeguards against new risks such as hidden jailbreak attacks that use interfaces and communications with agents to circumvent established restrictions, noting that defenses remain at an early stage globally. The agency is also using AI agents within Qainar, its collective cyberthreat protection system connecting 100 financial institutions, to analyze and structure cyber incident information and support rapid threat intelligence sharing. It is expanding the use of its own AI agents to monitor certain bank systems.
Agency for Regulation and Development of the Financial Market of the Republic of Kazakhstan outlines end-to-end oversight priorities for autonomous AI agents
Kazakhstan’s financial regulator outlined an end-to-end approach to supervising AI agents, centered on transparency, human control and customer protection. It called for safeguards against autonomous-agent risks, including hidden jailbreak attacks. The agency is also expanding supervisory AI use and operates AI agents within the Qainar cyber defense system, which connects 100 financial institutions.