AI Governance Institute
← News

Treasury Nominee Leaves Agentic AI Bank Liability Unresolved at Senate Hearing

What happened

At a November 20, 2026 Senate Banking, Housing and Urban Affairs Committee hearing, Rebekah Jurata, nominated for assistant secretary at the Treasury Department, was asked directly who should be held responsible if a third-party AI agent made an unauthorized transfer in a consumer bank account. As reported by FedScoop in Treasury nominee noncommittal on responsibility for rogue AI agents in accounts, Jurata cited 'further development of technology' and 'a lot of unanswered questions,' declining to take a position. She committed only to coordinating with other bank regulators within the existing legal framework. Separately, the Bank Policy Institute published a post the same day stating that risk management for agentic AI is 'not sufficiently developed,' and that agents should not yet be granted access to financial accounts. The Institute specifically identified prompt injection — a technique where malicious instructions hidden in content an agent reads can redirect the agent's actions — as a threat capable of causing accounts to be drained, and said current AI models are 'not nearly good enough' at resisting it. This hearing follows an earlier September statement by Treasury Secretary Scott Bessent, who told a House Financial Services Committee that AI creators should be held 'liable for what they build and generate,' adding a conflicting signal to an already unclear regulatory picture.

Why it matters

  • ·Financial services firms operating in the US have no regulatory guidance on where liability lands when an AI agent causes unauthorized account activity. Without that clarity, compliance teams cannot rely on regulatory frameworks to allocate responsibility and must address it contractually with vendors and operationally through internal controls.
  • ·The Bank Policy Institute's finding that agentic AI risk management is 'not sufficiently developed' for account access is a named industry risk signal. Combined with the prompt injection threat it identified, firms that have already granted third-party AI agents account-level access face unquantified exposure that existing model risk frameworks may not fully capture, as noted in Financial model risk frameworks miss most AI systems.
  • ·The hearing adds to a pattern of regulatory ambiguity in financial services AI: five major US banks have disclaimed AI agent liability, Bank of America warned customers that sharing banking passwords with AI agents voids fraud liability, and examiners are now asking banks to prove AI agents can be audited. Compliance teams should treat the absence of a Treasury position as a prompt to build their own liability allocation framework now.

Stay ahead of stories like this

Get every US AI governance development like this one, plus the rest of the week's developments. Every Thursday.

Free every Thursday. Unsubscribe anytime.

Governance controls affected

What to do now

  • ☐Identify every third-party AI agent or tool your organization has granted access to financial accounts or payment systems, and document what actions each agent can take without human approval.
  • ☐Review vendor contracts for AI agents with account access to confirm whether your organization or the vendor bears liability for unauthorized agent actions, and escalate gaps to legal counsel for renegotiation.
  • ☐Ask your engineering or vendor risk team whether the AI agents you use have been tested against prompt injection, specifically in scenarios where the agent reads external content such as emails, documents, or web pages before taking financial actions.
  • ☐Require human approval for any AI agent action that initiates, modifies, or cancels a financial transaction, and document the approval gate in your model risk or agentic AI governance program.
  • ☐Brief your compliance committee on the Bank Policy Institute's statement that agentic AI risk management is not sufficiently mature for account access, and assess whether your current deployment posture is consistent with that industry position.

What to watch next

Congress is actively pressing Treasury on agentic AI accountability, and the Senate Banking Committee hearing signals this will not be resolved quickly through voluntary coordination. Compliance teams should monitor whether Treasury issues formal guidance under the us-treasury-ai-risk-management-framework-financial-services-2026 or whether bank examiners begin incorporating agentic AI account-access questions into supervisory reviews, a trend already visible in banking examiners asking banks to prove AI agents can be audited. The Interagency Revised Guidance on Model Risk Management may also evolve to address agentic systems explicitly. Firms should also track whether Treasury Secretary Bessent's stated position that AI creators should bear liability for what they build gains traction in forthcoming legislation.

Related Coverage

Corporate Policy2026-11-19

Five Major US Banks Disclaim AI Agent Liability, Including After Hacking

Citibank, JPMorgan, Bank of America, PNC, and Citizens have each updated their personal account terms of service. The updates hold customers solely responsible for losses caused by third-party AI agents granted account access. Citibank's policy goes further, placing responsibility on customers even when the AI tool is hacked and their credentials are stolen by attackers. The updates affect personal account holders who authorize AI tools to access their accounts, not enterprise or business banking accounts.

Standards2026-11-16

Microsoft Defense Guide Makes Layered Prompt Injection Controls an Enterprise Baseline

Microsoft has published updated guidance on defending against indirect prompt injection, a category of attack where malicious instructions hidden in external content redirect AI systems to perform unintended actions. The guidance, titled 'Defend against indirect prompt injection attacks,' recommends a layered set of controls including prompt shields, content isolation, behavioral monitoring, and human approval gates. It applies to any organization using AI systems that process third-party content such as emails, documents, websites, or plugins.

Research2026-11-13

EchoLeak Proves AI Assistants Cannot Defend Against Prompt Injection Alone

A flaw named EchoLeak was patched by Microsoft in May 2025. It showed that a crafted email could cause Microsoft Copilot to silently send sensitive data to an attacker through an automatically loaded image. Security researchers publishing in November 2026 argue the underlying weakness persists across all AI assistants, because no AI model can reliably distinguish legitimate content from hidden attacker instructions. Enterprises using AI assistants with access to email, documents, or internal systems need architecture-level controls that sit outside the model itself.