US regulators probe banks over agentic AI, Chinese‑made AI models and autonomous system safeguards
agents autonomous
| Source: Techmeme | Original article
U.S. banking regulators are probing banks on their use of agentic AI, Chinese-developed AI models, and safeguards for autonomous systems as they intensify AI risk oversight.
U.S. banking regulators have begun probing financial institutions about how they are using advanced “agentic” artificial‑intelligence systems, whether they rely on AI models developed in China, and what safeguards are in place for autonomous decision‑making tools, six people familiar with the inquiries said.
The questioning marks a sharpening of supervisory focus on AI as banks integrate increasingly sophisticated models into credit underwriting, fraud detection and customer service. Regulators are concerned that agentic AI—systems capable of initiating actions without direct human prompts—could amplify operational risk, create new avenues for regulatory non‑compliance, or expose firms to geopolitical supply‑chain vulnerabilities if Chinese‑origin models are involved. The move follows a broader regulatory push to contain AI‑related threats across sectors, as highlighted in recent industry reporting.
What makes the effort noteworthy is its breadth: rather than targeting a single application, supervisors are asking banks to map the full lifecycle of AI use, from model sourcing to monitoring of autonomous outputs. The inquiries also signal that U.S. authorities may soon formalise expectations around transparency, testing and incident‑response procedures for AI‑driven processes.
Stakeholders should watch for any forthcoming guidance or rulemaking that codifies these expectations. Further, banks may need to disclose AI‑related risk assessments in their periodic filings, and could face heightened supervisory scrutiny if gaps are identified. The next weeks are likely to reveal whether the regulators will issue formal directives or keep the pressure informal, shaping how the financial sector balances innovation with risk containment.
Sources
Back to AIPULSEN