Financial services firm Equals Money PLC is opening Model Context Protocol (MCP) servers to customer AI tools with strict operational boundaries. Speaking at Okta's Oktane event, Chief Operations and Product Officer James Simcox announced that the company limits customer-facing AI agent access to data retrieval and non-sensitive updates, explicitly prohibiting agents from initiating money transfers.
Treating AI Agents Like Human Staff
To mitigate risk in payment processing, Equals Money treats software agents under the same audit and identity controls applied to human employees. Simcox noted that software agents require distinct identities and explicit access revoking mechanisms because a rogue agent deployment can scale to thousands of problematic instances instantly.
Internal adoption of AI tools at Equals Money is already substantial, with Simcox estimating that artificial intelligence generates roughly 92% of the company's code.
Automated Revocation and Security Controls
Equals Money currently handles agent token cancellation through manual intervention, but the company is looking to adopt automated detection systems to immediately terminate anomalous agent activity. Okta recently introduced runtime enforcement alongside an Agent Gateway kill switch aimed at invalidating all active tokens associated with a compromised agent.
While integrations with external tools such as enterprise resource planning software allow faster data exchange via MCP servers, Equals Money operates under the assumption of full liability for customer-side errors, prompting its ongoing restriction against autonomous payment execution.
AI-generated editorial illustration; not a documentary photograph.
Source: Equals Money lets customers’ AI tools read data but not move money
SIGNAL TITANS
Leave a Reply