Financial technology corporation Fidelity National Information Services has announced the commercial release of its new AML Investigations Agent. Developed in partnership with artificial intelligence firm Anthropic, the system is designed to automate the detection, investigation and documentation of illicit financial flows within financial institutions. The software runs as an autonomous agent in a private cloud environment to satisfy rigorous banking compliance and data protection mandates. The move addresses an acute industry problem, as financial institutions struggle with surging alert volumes and expensive manual review queues.
Initial operational metrics were gathered in collaboration with pilot partner Amalgamated Bank, which tested the architecture ahead of its general commercial rollout. According to FIS, the Claude-based agent slashed the review duration for standard Level-1 money laundering alerts by more than 90 percent. During deeper Level-2 investigations, the recorded time savings reached 98 percent. Previously, compliance departments relied heavily on human analysts to triage vast backlogs of false alarms generated by legacy rule-based software.
Beyond initial alert processing, the software automates the generation of regulatory filings required by law enforcement. The creation time for Suspicious Activity Report narratives, which document complex transaction patterns, decreased by more than 80 percent. Human analysts remain legally responsible for final sign-offs, but they now review pre-compiled dossiers with contextual transaction histories rather than assembling data manually. This operational change shifts AML workloads away from rote investigation toward final oversight and forensic sign-off.
The commercial rollout coincides with heightened regulatory scrutiny surrounding autonomous agentic systems. On the same day, the Financial Stability Board published its final guidance document outlining twelve recommendations for artificial intelligence adoption in financial services. The international body explicitly warned against unpredictable emergent behaviors in autonomous agents that could trigger transaction errors or broader market instability. The FSB standards require bank boards to maintain strict end-to-end accountability across the entire model life cycle.
To satisfy compliance regulators, foundational AI providers are building specialized verification frameworks. Anthropic engineers describe this safeguard as a multi-tier trust architecture, where agents synthesize evidence and draft resolutions, but execution requires human cryptographic signatures. Similar protective measures are being adopted across Wall Street, with quantitative firms like Two Sigma and financial data providers like Bloomberg enforcing deterministic input boundaries and automated classification filters. For banks, these technical barriers determine whether autonomous systems will survive strict regulatory audits.