Banks are running artificial intelligence at enormous scale. Federal regulators have just gone out of their way to say that the newest and most powerful category of that technology sits outside their model risk rulebook. This week Mike and Laura unpack that tension and what it means for anyone accountable for governance in a regulated institution.
AI at production scale. Bank of America reports more than three hundred approved AI and machine learning use cases, including one hundred fourteen live generative AI use cases with thirty four fully deployed. On the second quarter earnings call, CEO Brian Moynihan said more than two hundred thousand employees use AI enabled capabilities generating over four hundred thousand prompts daily, as reported by Banking Dive and CIO Dive. Jamie Dimon put JPMorgan Chase at almost one thousand AI use cases, and Jane Fraser said nearly nine out of ten Citigroup employees are using the bank’s AI tools.
The deliberate carve out. On April 17, 2026 the OCC, the Federal Reserve Board and the FDIC jointly issued revised interagency guidance on model risk management, superseding the 2011 guidance. The revised guidance states that generative AI and agentic AI models are novel and rapidly evolving and are not within its scope. The agencies said they plan to issue a request for information addressing model risk management generally and considering, in particular, banks’ use of AI. The guidance is expected to be most relevant to banking organizations with over $30 billion in total assets, and the OCC was explicit that it sets no enforceable standards and that non compliance will not result in supervisory criticism.
Supervision follows the same logic. On August 27, 2026 the OCC and the FDIC issued a final rule establishing a uniform definition of unsafe or unsound practice, directing examiners to prioritize material financial risks over policies, process and documentation, and setting a uniform standard for Matters Requiring Attention. An MRA now requires a practice contrary to generally accepted standards of prudent operation that could materially harm the bank’s financial condition or present a material risk of loss to the Deposit Insurance Fund, or an actual violation of law.
And a new competitor. In August 2026 the FDIC approved the deposit insurance application for Augustus National Bank, a proposed national bank in Dallas, Texas built to serve digital asset companies, artificial intelligence companies, technology companies, high net worth individuals and international financial institutions, with a tier one leverage ratio condition of no less than ten percent through its first three years.
The through line: discretion has moved back to the institution. An empty rulebook is not permission. It is an invitation to write the standard yourself and be ready to show your work.
To learn more about PiTech Solutions and how we help regulated institutions build governance that holds up under examination, visit pitechsol.com.