Work
AI provenance and MLS compliance
The subsystem that lets AI-altered listing photos go out legally — watermarking, C2PA provenance, and export gates that know what state a listing is in.
- Period
- 2026
- Role
- Designed and built
- Stack
- C2PA · AB 723 / SB 942 · Next.js
- Result
- 4 compliance regimes covered
California AB 723 and SB 942, the NAR Code of Ethics, and MLS rules all landed on generative real estate imagery at roughly the same time, and they don’t say the same thing. Shipping staged photos without a compliance layer stopped being viable.
What it does
- Per-image watermarking on every generated asset.
- QR-linked original preservation, so anyone looking at a staged photo can reach the unmodified one.
- Listing-state-aware export gates — what you’re allowed to export depends on where the listing is in its lifecycle, so the rule lives in the system rather than in an agent’s memory.
- Server-side attestation logging, because the defensible position is a log you didn’t write after the fact.
- C2PA provenance embedded in the asset itself.
Why it’s here
Compliance subsystems are unglamorous and rarely make portfolios. This one is on the list because it’s the part of the product that determines whether a brokerage can adopt it at all — legal review is the gate, not the demo.