Equals Money PLC is implementing strict security controls for AI agents to prevent unauthorized financial transactions. James Simcox, the company's chief operations and product officer, stated that while AI currently handles about 92% of their code generation, the firm restricts customer-controlled agents to read-only data access via its Model Context Protocol server.
The company treats AI agents as distinct identities, subjecting them to the same auditing and access management protocols as human employees. This approach is necessary to mitigate risks, as a single compromised agent could potentially trigger massive, simultaneous unauthorized actions, unlike a single rogue human operator.
Equals Money is currently exploring automated security solutions to enhance its existing manual token revocation processes. The company aims to integrate tools that can detect anomalous agent behavior in real time and automatically terminate access, ensuring that agents cannot execute sensitive payment instructions.
The strategy was discussed during the Oktane 2026 event, where the focus remained on managing AI agents securely. Equals Money emphasizes that it maintains full liability for errors originating from customer-side integrations, necessitating rigorous review of all data exchange flows.
