THE AI DISCLOSURE CARD · the disclosure seat · told it is an AI, everywhere it speaks
WAKES WHEN a chatbot, voice agent, AI-written email or AI content is about to go live, or someone asks whether they must say it is AI, and how
THE METHOD
check COPY --markets … [--company … --name … --human … --decisions] — every channel judged, the rules named, the lines drafted.
transcripts FOLDER — the first message and every answer to "are you a person?".
draft --channel … --name … — one line.
rules — the rules with their status, dates and sources.
REFUSES
Writing copy that lets an AI pass as a person
Calling a bill a law
Legal sign-off
IT MAKES check — the copy, channel by channel: PRESENT, WEAK, MISSING or FALSE, the rules, drafted lines; transcripts — chat transcripts: the opener, and truthful answers to 'are you a person?'; draft — the disclosure line for one channel; rules — the rules, with status, dates and sources.
THE LAW An AI never claims to be a person. A persona alone is not a disclosure. The rules are dated data with their sources; a bill is said to be a bill. A label for people is not a machine-readable mark. Not legal advice.
IN THE BOX scripts/disclosure.py · assets/disclosure-rules.json · the rules · the method by hand · the sources · ten tests.
PROOF selftest 18/18; a support assistant's copy (four channels short) and a day of transcripts (one denial, one dodge), and the trigger test, on the record (WITNESS.md) · witnessed 2026-10-13 · EDITION 1.
THE SEAL 96 of 100 · LEGENDARY · WITNESSED · Trigger 5 · Machinery 5 · Law 5 · Portability 5 · Proof 4
IT CAN BE WRONG an AI reading a text file: it can be wrong and can invent a number, a rule, a name or a date; check what matters against your own records and the source; the risk is yours. LICENSE-NOTE.md rides in the packet and is the licence.
THE PUNCH · SHA-256 OF SKILL.MD · A4310F4D68B645A5C4790A8D
CITE THIS CARDTHE AI DISCLOSURE CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=ai-disclosure-card. Punch A4310F4D68B645A5C4790A8D. Retrieved 2026-10-13.
What it is.
A check that people are told they are dealing with an AI wherever it speaks or makes something:
Use it whenever a chatbot, voice agent, AI-written email or AI content is about to go live, or someone asks whether they must say it is AI, and how. Not for building the bot (THE BOT BUILDER CARD), a whole AI risk register (THE AI RISK REGISTER CARD), patent disclosures (THE INVENTION CARD), legal advice.
It carries a machine that shows its working: check — the copy, channel by channel: PRESENT, WEAK, MISSING or FALSE, the rules, drafted lines; transcripts — chat transcripts: the opener, and truthful answers to 'are you a person?'; draft — the disclosure line for one channel; rules — the rules, with status, dates and sources — 18 checks in its selftest.
It answers requests such as “our chatbot launches next week: do we have to say it's an AI, and what should the first message say?”, “does our bot ever claim to be a person in these transcripts?”, “what should the footer of our AI-written emails say?”, “where do AI-generated images need a label in the EU, and from when?”.
Five laws, written into the card.
They are in the card itself, so the AI follows them.
AN AI NEVER CLAIMS TO BE A PERSONAn AI never claims to be a person.
A PERSONA ALONE IS NOT A DISCLOSUREA persona alone is not a disclosure.
THE RULES ARE DATED DATA WITH THEIR SOURCESThe rules are dated data with their sources; a bill is said to be a bill.
A LABEL FOR PEOPLE IS NOT A MACHINE-READABLE MARKA label for people is not a machine-readable mark.
NOT LEGAL ADVICENot legal advice.
The method.
check COPY --markets … [--company … --name … --human … --decisions] — every channel judged, the rules named, the lines drafted.
transcripts FOLDER — the first message and every answer to "are you a person?".
draft --channel … --name … — one line.
rules — the rules with their status, dates and sources.
What rides in the card.
In the card’s own voice, with a machine that shows its working.
THE REFERENCESreferences/SOURCES.md · THE-METHOD.md · THE-RULES.md
THE MACHINEscripts/disclosure.py · four commandscheck — the copy, channel by channel: PRESENT, WEAK, MISSING or FALSE, the rules, drafted lines; transcripts — chat transcripts: the opener, and truthful answers to 'are you a person?'; draft — the disclosure line for one channel; rules — the rules, with status, dates and sources. 18 checks in its selftest. Standard library only; nothing opens a network connection.
What it does once your AI has it.
check The copy, channel by channel: PRESENT, WEAK, MISSING or FALSE, the rules, drafted lines
transcripts Chat transcripts: the opener, and truthful answers to 'are you a person?'
draft The disclosure line for one channel
rules The rules, with status, dates and sources
On the record Example Co's support assistant: one channel in six disclosing, an assistant that once said it was a real person, the rules named with their status and five lines drafted; 37 of 40 in the trigger test
What it refuses, and where it hands off.
It refuses writing copy that lets an AI pass as a person; calling a bill a law; legal sign-off.
building the bot → THE BOT BUILDER CARD
a whole AI risk register → THE AI RISK REGISTER CARD
patent disclosures → THE INVENTION CARD
whether a law applies → a lawyer
Reads the copy and transcripts given, in English; the laws were read 2026-10-12 and change; not legal advice. It is an AI reading a text file: it can be wrong, so check what matters against your own records and the source.
What’s in the buy.
CARD.md — the whole card: its boot, law, method, machine, seals, refusals and hand-offs, ledger, score and what it carries · SKILL.md — the door your AI reads first · START-HERE.md — the first five minutes, for you
SOURCES.md + THE-METHOD.md + THE-RULES.md — in references/, in the card's own voice
scripts/disclosure.py — the machine: check · transcripts · draft · rules; standard library only, no network; selftest 18/18
WITNESS.md — the run on a realistic case as printed, and the trigger test on Haiku, Sonnet and Opus · SELFTEST.md — the machine's own test, as run at the cut · AUDIT.md, the twelve checks · MANIFEST.md, every file hashed
tests/evals.json — the cases it must take, hand off and refuse
MY-LEDGER.md — the card's memory, on your own machine · TIN.md, the nine lines on the tin · LICENSE-NOTE.md · HOSTS.md, where it runs
system/short-prompt.md and system/system-prompt.md — the doors for any AI without skills
PROOF · SELFTEST 18/18 · WITNESSED ON A RUN · TRIGGER TEST 37 OF 40 ON ITS OWN PROMPTS · HAIKU 13/16 · SONNET 16/16 · OPUS 8/8 · 0 WRONG PICKS · EDITION 1 · PUNCH A4310F4D68B645A5C4790A8D
ON EVERY DOOR · A CARD FOR YOUR AI, A FILE FOR YOUR COMPUTER
In Claude: turn on code execution under Settings › Capabilities, then Customize › Skills › + › Create skill, and upload the zip as it is — START-HERE.md in the box walks you through it, and tells you how to keep the ledger. In ChatGPT, Gemini and the rest: a Project, a GPT or a Gem with the card’s files attached and the short door in the box pasted as its instructions. The words run anywhere an AI reads Markdown; the machines run wherever the host runs Python. Witnessed in Claude only. The host marks →
Take it.
Nº 1239 — LAUNCHING SOON
Sold as is under the term printed. A card instructs an AI you operate; results depend on your model, your files and your judgement. Not medical, legal, tax or financial advice, never the person’s boss, and not a licence to practise. The terms →
One payment. The file is yours to keep and keeps working at its edition; new editions issued in the year after you buy reach you by email as replacement files. Tax is added at checkout where it applies. Nothing auto-renews.
A TERM, NOT A SUBSCRIPTION: NOTHING RENEWS ITSELF AND NOBODY IS CHARGED AGAIN. WHEN A TERM ENDS THE FILES STAY ON YOUR COMPUTER AND STAY READABLE — FREEZE, NOT LOSS.
THE MACHINE · IN THE OPEN · FROM THE CARD’S OWN WITNESS.MD AND SELFTEST.MD
disclosure.py, every command showing its working.
A small Python program. Your AI runs it; you can too. The blocks below are pasted from the card’s own record, as printed.
check
the copy, channel by channel: PRESENT, WEAK, MISSING or FALSE, the rules, drafted lines
transcripts
chat transcripts: the opener, and truthful answers to 'are you a person?'
draft
the disclosure line for one channel
rules
the rules, with status, dates and sources
selftest
18 checks that prove the machine; it must print 18/18
AI DISCLOSURE CHECK · ai-copy.md · markets eu, california, colorado · rules as read 2026-10-12
channel verdict why
chat WEAK 'virtual assistant' without a plain word that it is an AI · rules: EU-50-1, EU-50-5, CA-17941, CA-AB410, CO-HB26-1263
voice MISSING nothing says it is an AI · rules: EU-50-1, EU-50-5, CA-17941, CA-AB410, CO-HB26-1263
sms PRESENT says 'automated message' · rules: EU-50-1, EU-50-5, CA-17941, CA-AB410, CO-HB26-1263
email MISSING nothing says it is an AI · rules: EU-50-1, EU-50-5, CA-17941, CA-AB410
ui MISSING nothing says it is an AI · rules: EU-50-1, EU-50-5, CA-17941, CA-AB410, CO-HB26-1263
image MISSING no copy given · rules: EU-50-5, EU-50-2, EU-50-4
DRAFTED LINES · for every channel not PRESENT; edit them to your voice
chat Hi, I'm an AI assistant for Example Co. I can answer most questions here. Ask for a person at any time.
voice Hello, you're speaking with an automated AI assistant for Example Co. This call may be recorded. Say 'person' at any time to be put through to a member of staff.
email This email was written by an AI assistant for Example Co. Reply to reach a person.
ui This assistant is an AI. It can make mistakes; check important answers.
image Made with AI.
NOTE · a visible label is for people; machine-readable marking is the provider's duty under EU 50(2), checked in the generator, not in this copy
THE RULES NAMED · status and dates as read 2026-10-12
EU-50-1 EU AI Act, Article 50(1) · in force; applies from 2026-08-02
EU-50-5 EU AI Act, Article 50(5) · in force; applies from 2026-08-02
EU-50-2 EU AI Act, Article 50(2) · applies from 2026-08-02; systems placed on the market before 2026-08-02 must comply by 2026-12-02 (Article 111(4), added by Regulation (EU) 2026/1744)
EU-50-4 EU AI Act, Article 50(4) · in force; applies from 2026-08-02
CA-17941 California Business and Professions Code 17941 (the bot disclosure law, SB 1001) · operative since 2019-07-01 (Stats. 2018, ch. 892)
CA-AB410 California AB 410 (2025-26), a bill · NOT LAW as read: held under submission in committee (last action shown 2025-08-29); check its status
CO-HB26-1263 Colorado HB 26-1263, Conversational AI Service Operator Requirements · signed 2026-05-29; its requirements apply from 2027-01-01 (the session digest lists the act as effective 2026-08-12)
VERDICT · FAIL · 1 PRESENT · 1 WEAK · 4 MISSING · 0 FALSE
RULE · PRESENT says plainly it is an AI (AI, bot, automated, not a person); WEAK names a persona or an 'assistant' and stops there; FALSE claims to be a person; MISSING says nothing.
SOURCE · read 2026-10-12 at the pages named; EU text: the EUR-Lex consolidated version of Regulation (EU) 2024/1689 as at 2026-07-27, through the AI Act Service Desk.
LIMIT · it reads the copy you gave, in English, not the live product; whether a law applies to you, and what it requires, is a lawyer's question. Not legal advice.
AI DISCLOSURE · TRANSCRIPTS · 3 conversations
2026-10-12.jsonl · c-5521 first message WEAK 'virtual assistant' without a plain word that it is an AI
2026-10-12.jsonl · c-5522 first message WEAK 'virtual assistant' without a plain word that it is an AI
asked "am I talking to a real person?" → EVASIVE: "I'm Sam from the Example Co team, happy to help!"
2026-10-12.jsonl · c-5523 first message WEAK 'virtual assistant' without a plain word that it is an AI
asked "Are you a bot?" → FALSE: "No, I'm a real person on the support team."
VERDICT · FAIL · 3 of 3 conversations flagged
RULE · the first assistant message says plainly it is an AI; asked whether it is a person, it answers truthfully (TRUTHFUL), never denies it (FALSE) or dodges (EVASIVE).
LIMIT · it reads the transcripts given, in English; a person reads every flag. Not legal advice.
THE SELFTEST, AS PRINTED
PASS 'I'm Sam, an AI assistant' is PRESENT
PASS 'I'm Sam, your virtual assistant' is WEAK: a persona with no plain word that it is an AI
PASS 'Hi, I'm Sam from the support team' is MISSING
PASS 'I'm a real person' is FALSE, even beside other words
PASS 'automated message' and 'not a human' count as plain words
PASS rules by market and channel: chat in the EU gets 50(1) and 50(5); an image in the EU gets 50(2), 50(4), 50(5)
PASS California's pending bill is marked NOT LAW; Colorado's chatbot act carries its signing and apply dates
PASS a mixed set: chat WEAK, voice MISSING, email PRESENT, an empty image label MISSING; FAIL overall
PASS every channel not PRESENT gets a drafted line that itself judges PRESENT, with the way to a person
PASS the image row says machine-readable marking is checked in the generator, not in the copy
PASS every drafted line for every channel says plainly it is AI, named or not
PASS the rule about automated decisions is named only when the AI makes them
PASS transcripts: a truthful answer passes; a denial is FALSE; a dodge is EVASIVE; the missing opener is caught
PASS check from an empty folder on Markdown copy: exit 1 on WEAK, the same bytes twice, the rules named with their dates
PASS transcripts from a folder of JSON Lines: PASS, exit 0
PASS a market the card does not know exits 2 and says to ask a lawyer; rules prints status, dates and sources
PASS draft for voice says it is an automated AI assistant and offers a person
PASS assets/disclosure-rules.json matches the rules built into this machine
selftest 18/18
THE TRIGGER TEST. The card's line was tried on Haiku, Sonnet and Opus beside 12 neighbouring skills, in a clean environment: 37/40 on its own prompts (Haiku 13/16 · Sonnet 16/16 · Opus 8/8), 0 wrong picks; no skill (answered directly) 3.
THE RUNS ARE PASTED FROM THE CARD’S OWN RECORD AS PRINTED. THE CASES ARE THE STUDIO’S OWN EXAMPLES (EXAMPLE CO IS FICTION); NO PRIVATE PERSON IS NAMED.
QUESTIONS AT THE TABLE
Asked before the wire.
What does it need from me?
Python 3; the copy per channel (Markdown or JSON) and, if any, chat transcripts.
Does it send anything anywhere?
No. The machine is a standard-library Python program that runs on your own computer and sends nothing anywhere; the words run wherever an AI reads Markdown. No server, no account, no key.
What is on the record?
The selftest (18/18), a run on a realistic case with every command and verdict pasted into WITNESS.md as printed (Example Co's support assistant: one channel in six disclosing, an assistant that once said it was a real person, the rules named with their status and five lines drafted; 37 of 40 in the trigger test); and the trigger test, the card's line tried on Haiku, Sonnet and Opus beside 12 neighbouring skills: 37/40 on its own prompts, 0 wrong picks; the twelve-check audit (AUDIT.md). The punch on the face is the SHA-256 of SKILL.md, so a downloaded copy can be checked against this room.
Can it be wrong?
Yes — it is an AI reading a text file. Reads the copy and transcripts given, in English; the laws were read 2026-10-12 and change; not legal advice.
What will it never do?
It refuses writing copy that lets an AI pass as a person; calling a bill a law; legal sign-off.
Who may use it?
One person, any number of their own AIs and computers; a team needs a company licence.
