The Federal Reserve named the gap.
GRACE closes it.
SR 26-2 supersedes SR 11-7 as the Federal Reserve’s primary AI supervision guidance. Its Footnote 3 is the most consequential sentence for institutions deploying agentic AI — it excludes generative and agentic AI from formal scope while simultaneously requiring institutions to self-govern.
This creates an examination requirement with no defined evidence standard. GRACE’s Policy Action Packet architecture is the evidence standard Footnote 3 implicitly requires.
The agencies have also announced they plan to issue a request for information on model risk management and banks’ use of AI — including generative and agentic AI — in the near future. GRACE is positioned ahead of that RFI.
SR 26-2 · Footnote 3 · Verbatim
“Generative AI and agentic AI models are novel and rapidly evolving. As such, they are not within the scope of this guidance. Nonetheless, a banking organization’s risk management and governance practices should guide the determination of appropriate governance and controls for any tools, processes, or systems not covered in this document.”
Federal Reserve SR 26-2, April 17, 2026. Issued jointly by Fed, OCC, and FDIC.
Filed with the agencies responsible
for defining the standard.
Federal Reserve
June 2026
GRACE submitted a regulatory comment letter and Technical Architecture Reference to the Federal Reserve, proposing the PAP architecture as the pre-execution enforcement layer referenced but not specified in SR 26-2 Footnote 3.
White paper submitted to Federal Reserve, June 2026.
OCC
June 2026
Grace AI Control submitted a formal technology introduction positioning GRACE as the candidate reference architecture for the forthcoming interagency agentic AI RFI.
White paper submitted to OCC Office of Innovation, June 2026.
FDIC
June 2026
GRACE’s FDIC submission documents how the PAP architecture addresses third-party AI risk, model risk, and the evidentiary requirements of FDIC technology supervision examinations.
White paper submitted to FDIC Division of Risk Management Supervision, June 2026.
FFIEC
June 2026
Grace AI Control submitted a formal architecture candidacy to the FFIEC Technology Working Group, proposing GRACE as a candidate framework for the forthcoming agentic AI governance examination section.
Architecture candidacy submitted to FFIEC Technology Working Group, June 2026.
Regulators described GRACE’s architecture in their own words.
FINRA’s 2026 Annual Regulatory Oversight Report classifies AI agents as a distinct supervisory risk category and recommends firms implement:
Narrow scope and explicit permissions for each agent
Complete audit trails of all agent actions
“Human checkpoints before execution”
“Human checkpoints before execution” is the GRACE architecture described in FINRA’s own language.
Provisional patent No. 63/960,895 filed. Priority date established.
Cloud Security Alliance paper: “Uneven Resource Exhaustion in AI Systems.”
SR 26-2 issued. Footnote 3 confirms the enforcement gap GRACE was built to close.
Axios reports $500M AI cost overrun — the exact failure mode documented in the February paper.
White papers filed with Federal Reserve, OCC, FDIC. OFT and FFIEC submissions filed.
Non-provisional patent filing in progress with Donald J. Ersler.
Interagency agentic AI RFI anticipated. GRACE positioned ahead of comment window.