Regulatory Updates Using AI Chat and AI Projects

A guide to using AI to update your Policies and Procedures — when to use a Chat and when to use a Project. Chat mode handles one regulatory update against one P&P; Project mode handles a regulation that reaches several P&Ps at once. Same goal, picked by scale. Every regulatory update below is sorted into the way it’s best run, in calendar order.

Chat mode

One regulatory update, one P&P

What you’ll upload to the chat window: the regulatory update in a machine-readable format — CSV or Markdown — plus the P&P you want to update. Ask the AI to tell you where your P&P isn’t in line with regulatory requirements — it matches the sections, checks alignment, and hands you a redline and a gap analysis.

Try it on the FHA Loss Mitigation update below. It’s already in effect, so it’s a chance to double-check your work: open the machine-readable file, drop it in a chat with your FHA Loss Mitigation P&P, and use our prompt to compare the two. The prompt and the machine-readable file work in tandem to combat hallucinations.

Worked example: the FHA Loss-Mit kit →

  1. Oct 1in effect
    AVM Quality-Control Rule (12 CFR 34.222)
    Open the kit →
  2. Feb 2in effect
    FHA Loss Mitigation Overhaul — Mortgagee Letter 2025-06
    Open the kit →
  3. Mar 4in effect
    HPPA - FCRA Trigger-Lead Restriction
    Open the kit →
  4. Apr 29in effect
    FHA Foreclosure Bidding & CWCOT (ML 2026-03)
    Open the kit →
  5. Jun 1in effect
    VA Partial Claim Program (38 U.S.C. 3737)
    Open the kit →
  6. Jun 17in effect
    USDA SFHGLP - Delegated Authority Final Rule
    Open the kit →
  7. Jun 23in effect
    FHA Limited 203(k) - Draw Request Limits (ML 2026-06)
    Open the kit →
  8. Jun 23in effect
    FHA - Rescission of the Important Notice to Homebuyers (ML 2026-07)
    Open the kit →
  9. Jun 23in effect
    FHA - Streamlined Mortgagee-Approval & QC (ML 2026-09)
    Open the kit →
  10. Jun 23in effect
    FHA QC - Appraisal Field Reviews (ML 2026-10)
    Open the kit →
  11. Sep 1upcoming
    USDA Section 502 LITE Delegated-Authority Pilot
    Open the kit →
  12. Sep 21upcoming
    FHA Loss Mitigation - Trial Payment Plans & Foreclosure Timing (ML 2026-08)
    Open the kit →
Project mode

One change, many P&Ps

A Project is a set of linked chats — one P&P per chat — that can compare notes across all of them. Upload every P&P into one Project and a single regulatory change becomes one coordinated review instead of forty separate errands.

The textbook case is the Fannie Mae AI Lender Letter (LL-2026-04): because AI touches so many P&Ps, you push the update to each one through Project mode. Our FNMA AI Lender Letter Survival Update Kit ships the machine-readable regulations plus the workflow to inventory the AI you already have, gather what’s missing, draft your AI master policy, and push the changes back to each individual P&P.

Worked example: the FNMA AI kit →

  1. Aug 6upcoming
    FNMA AI Lender Letter (LL-2026-04) — Survivors Compliance Guide
    Open the kit →

Is it really that easy? Yes and no. AI and large language models have hallucination problems, so the prompting and the regulatory files have to be built to guard against them. The machine-readable regulatory updates and Skills here carry tracking that ties every requirement to its verbatim source, so the AI’s work is checkable instead of trusted — the same anti-hallucination discipline behind every kit. If you have Codex, Cowork, or Claude Code, copy the methodology in, build a few of your own regulatory files, and have some learning-type fun. How we address hallucination in regulatory updates →

Compliance is everyone’s job, including Claude.

Claude is a tool that can increase accuracy and efficiency when the tool is used correctly. Learning more about Claude is win/win as you will understand how operations may be using Claude, and also how Claude can be used to facilitate compliance.

Compliance for Claude gives you mortgage compliance regulations saved as Claude files, so you can bring compliance work into your own Claude projects. Everything stays inside your company’s confidentiality protections.

If your journey looks like mine, you’ll start by using the Claude regulatory files to audit your P&Ps. Once you get started, it’s hard to stop, and the more you use Claude, the better your Claude/AI compliance program will be.

Here’s the whole idea, in four steps:

1
Open your AI chatClaude, Copilot, or ChatGPT
2
Upload the updated regulations FHA Loss Mit Regs .md
3
Upload your P&P SAMPLE Loss Mit P&P .md
4
Get your updated P&Predlined to the current rules

Those downloads are .md files — plain text (Markdown) that opens in Notepad and pastes straight into any AI chat. Nothing to install, nothing executable, just text you can read and edit.

These regulatory files aren’t just text. Each one carries tracking that ties every requirement to its verbatim source, so the AI’s work is checkable instead of trusted, the same anti-hallucination discipline behind every kit. And several of these rules are already in effect, so this isn’t hypothetical: it’s a chance to check your status and double-check your work against what’s already on the books. How we address hallucination in regulatory updates →

Use Claude’s compliance tracking, and supplement beyond it

Claude Compliance API is a must-use tool. It is the foundational audit record available. Claude Compliance API logs tool calls, and there is a full transcript available. You can set up additional tools in Claude to have Claude notate regulatory applicable workflow and actions in Claude, to understand what is happening, and be able to better answer regulatory questions.

Learn more about compliance tracking in Claude →

Our why

The more compliance professionals use AI, the better compliance will be with AI. Compliance teams are the ones who think about what happens when things go wrong, and every AI use case a firm wants to deploy needs that careful, cited, documented thinking done first. Compliance for Claude is built so compliance teams can do that thinking at the speed the firm needs, without giving up the rigor that makes the thinking worth doing. More on why →