Hearsay listens to regulators across every channel they use to speak.
Compliance teams monitor regulator websites and miss half the signal. Real regulatory intent surfaces first in hearings, then in PDFs, then on press pages — sometimes weeks apart. No GRC tool reads all three. Hearsay does.
Every regulator page fetch — JS-rendered, geo-targeted, anti-bot-resistant. SEC.gov, ESMA, MAS, BaFin and 121 more pull cleanly through one API.
PDFs and rendered screenshots — NIST CSF, CISA advisories, enforcement notices. We page-by-page render then ask the vision model what changed.
Hearings, press conferences, agency YouTube. Premium-tier classifier extracts forward-looking intent — the stuff that lands in a rule weeks later.
- 01Discover
Bright Data SERP enriches each regulator with its primary URL, press-release page, and hearings/media page. 125 regulators across 67 jurisdictions are seeded; bootstrap can scale to 500+.
- 02Fetch
Bright Data Web Unlocker pulls each page — JS-rendered when needed, geo-targeted to the regulator's home country. SEC.gov returns 403 on plain fetch; through Unlocker it returns 200.
- 03Extract
A cheap-tier model (gpt-4o-mini via AI/ML API) summarizes the page diff into discrete updates. Visual modality renders PDF pages to images and asks gpt-4o-vision what changed. Spoken modality runs Speechmatics over the audio and feeds the transcript to a premium tier.
- 04Classify
A mid-tier (gpt-4o) or premium tier (claude-opus-4-7) reasons over the update: severity, affected controls (SOC 2 / ISO / NIST), one-paragraph plain-English reasoning, three imperative recommendations. The model that ran is stamped on the card so failover is visible to the eye.
- 05Surface
Globe pings the jurisdiction. Alert lands in the wire with severity gutter, control tags, AI reasoning block, and 'What to mind' bullets. Click an alert to fly the camera; click a country to inspect its full feed.