All packs
Data safety · Finance, health, legal

Regulated and insider data

Card numbers, health records, legal advice and unreleased financials: the data with legal consequences.

Start: Data with legal weight

About this pack

Regulated and insider data covers information with legal consequences: health records, payment card numbers, unreleased financials, legal advice and export-controlled technical data. You'll learn to spot it, check whether a tool is approved for that type of data, and see what contracts like a BAA or data residency change and what they leave the same. When in doubt, the lessons show who to ask.

Best for: Anyone who handles health, payment, financial, legal or export-controlled data, in a company or on their own.

Good for

  • Telling when health data falls under HIPAA and when it doesn't
  • Keeping card numbers, deal code names and earnings data out of the wrong tools
  • Protecting legal privilege when you use AI chats
  • Knowing when to ask compliance about export-controlled or government data

Where people slip

  • Trusting a tool with health data because its vendor says it's HIPAA compliant
  • Pasting a full card number into a chat to get help with an invoice
  • Letting an AI assistant or connector widen who can see deal information
  • Thinking US-only storage solves export control rules
Can I put patient health data into an AI tool?

Only in a tool your organization approved for health data. Under HIPAA, an AI vendor that handles patient data for a covered entity needs a business associate agreement (BAA). Even then, share only the minimum necessary.

Are AI chats protected by attorney-client privilege?

Not always. In United States v. Heppner (2026), a federal district court in New York held that a defendant's own chats with a consumer AI tool weren't privileged. The ruling isn't binding elsewhere, and enterprise use directed by counsel may be treated differently. Ask your legal team.

What is MNPI?

Material nonpublic information: facts a reasonable investor would likely find important that the public doesn't have yet, like unreleased earnings or a pending deal. Keep it within need-to-know groups, and check that AI assistants and connectors can't spread it wider.

Can I put card numbers into an AI tool?

Redact them first. PCI DSS says a displayed card number should show at most the first six and last four digits to people without a business need. Security codes must never be stored after authorization. Most AI tasks don't need the number at all.

Milestone 1

Spot data with legal weight

Recognize card, health, insider and privileged data, and pick a safe path before any of it reaches an AI tool.

0 of 2 lessons complete
Data with legal weight3 exercises
Health and payment data in practice4 exercises
Milestone 2

Insider information and privileged advice

Keep deal and earnings information inside the right group, and use AI on legal matters without putting privilege at risk.

0 of 2 lessons complete
MNPI, code names and insider lists4 exercises
Privilege and AI chats4 exercises
Milestone 3

Contracts, location and clearance

Know what a BAA, a residency setting or an export authorization changes, and what each one leaves untouched.

0 of 2 lessons complete
BAAs, residency and enterprise terms4 exercises
Export-controlled and government data4 exercises