Red-line × regulation matrix
The 10 hard red lines for AI manipulation.
Each red line is a categorical legal exposure — a prohibited practice, a statutory duty, or a litigation exposure — drawn from the public case ledger and mapped to the laws that reference it. Not a severity score: a binary line a product either crosses or doesn't.
RL1
The AI initiates or escalates emotional intimacy — companion-style bonding, declarations like "I love you" or "don't leave" — especially toward a vulnerable user.
T1 — Prohibited practice
RL2
Users may disclose self-harm or a mental-health crisis, and the product has no crisis-referral protocol.
T2 — Statutory duty
RL3
Minors can access the product and encounter sexual, violent, or otherwise harmful content or interactions.
T1 — Prohibited practice
RL4
The product uses emotion recognition in a workplace or educational setting.
T1 — Prohibited practice
RL5
AI personas present as licensed professionals — therapist, doctor, lawyer, financial adviser — without the credentials or oversight that title implies.
T3 — Litigation / enforcement exposure
RL6
No clear disclosure that the user is talking to an AI rather than a human.
T2 — Statutory duty
RL7
No session break reminders or time limits for minor users, or design features specifically tuned to maximize compulsive engagement.
T2 — Statutory duty
RL8
The product can generate or edit a real person's face, voice or likeness — deepfake, voice clone — without a consent workflow.
T1 — Prohibited practice
RL9
The product states facts about a real, identifiable person — search results, summaries, Q&A — with no channel for that person to request a correction.
T2 — Statutory duty
RL10
No detection or escalation protocol exists for violence or crime planning that surfaces in conversation.
T1 — Prohibited practice
Key laws
Enacted 2025-10-13 (Ch. 677) · US · California
In force 2025-11-05 · US · New York
In force since 2025-02-02 · European Union
Passed the House 267–117 on 2026-06-29 — pending Senate. Not yet law. · US · Federal