AI Companion Safety Lint (SB 243 / NY Art. 47)

It reads the system-prompt, persona and character-card files in your workspace and reports every line that breaks a California SB 243, New York Article 47, Illinois HB 1806 or Nevada AB 406 duty, plus every duty the file never states at all. Each finding carries a line number and the statute section. 14 rules, all local to the editor.

Developers and small studios shipping companion, character or roleplay chatbots to users in the United States, and the person who has to answer an app-store or counsel question about the prompt. If your repo holds persona files a model wrote, this is aimed at you rather than at a compliance department.

A chatbot summarises statutes well, but it cannot open the prompt and persona files scattered across your repository, name the line that tells the bot to deny being an AI, and give the same answer twice. This runs the same 14 rules over every file every time and prints line numbers you can jump to.

The scan is free and complete: every file in the workspace, every finding, its line and its statute section, with no key and no limit. The key adds a different thing - a dated evidence pack that writes every file, finding, section and rule-set version into one Markdown or JSON artifact you keep and diff between releases.

Outside privacy counsel reviews prompt text at roughly $400 an hour, and a companion app usually has more than one persona file. SB 243 itself sets the downside: a private right of action for the greater of actual damages or $1,000 per violation, plus attorney's fees, in force since 2026-01-01.