THE INTRO CARD is free — what an AI skill card is and how to use one Download →Lost Lauren, contact us

THE INSTRUMENTS · Nº 1174 · EDITION 1 · LAUNCHING SOON

LUND CARDS · THE INSTRUMENTSNº 1174 PROPOSED
THE DEEPFAKE CHECK CARD — the LSC-1 face
THE CARD AS TEXT · FOR EYES AND FOR MACHINES

THE DEEPFAKE CHECK CARD · a file's content record, read before trust

WAKES WHEN someone is about to trust or share a photo, a voice clip or a video

THE LOOP

  1. inspect · the record, inside the file
  2. read what is there and what is missing
  3. verdict · one plain line a file
  4. folder · a whole folder, counted
  5. verify a record with a full validator

NEVER

  • Say a file is fake or real from its pixels
  • Call a record found a record verified
  • Read "no record" as proof of a fake

IT MAKES inspect — what is there, what is missing, the verdict; verdict — one line a file; folder — a folder, counted by verdict.

THE STANDARD the record, never the pixels; three plain verdicts; found is not verified; unsigned words, said so; no record proves nothing; damaged is cannot tell.

IN THE BOX SKILL.md · CARD.md · TIN.md · START-HERE.md · LICENSE-NOTE.md · THE-CLAUSES.md · MY-LEDGER.md · WITNESS.md · SELFTEST.md · AUDIT.md · MANIFEST.md · HOSTS.md · scripts/deepfakecheck.py · references/THE-CONTENT-RECORD.md · references/HORIZON-TWO.md · rules/ (4 doors) · system/ (2 prompts) · mcp/mcp_card.py · card.json · card.svg · card-machine.svg · card.html.

PROOF SELFTEST 8/8 · witnessed in Claude 2026-10-11 · Edition 1.

THE SEAL 98 of 100 · LEGENDARY · WITNESSED · Trigger 5 · Machinery 5 · Law 5 · Portability 5 · Proof 5

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. THE-CLAUSES.md rides in the packet and is part of the licence.

THE PUNCH · SHA-256 OF SKILL.MD · A5D242045764F08B82ADF7CD

THE FACE, FULL SIZE → · THE MACHINE FACE → · card.json →

THE DEEPFAKE CHECK CARD

a file's content record, read before trust

INSPECT · VERDICT · FOLDER

98LEGENDARYWITNESSEDIN CLAUDEIN LUNDRIN CHATGPT

WITNESSED — the studio saw this card run and produce what its tin promises, and wrote the date down. What the marks mean →

THE SPEC SHEET · READ BEFORE YOU BUY · THE STANDARD EVERY CARD MEETS →

SEAL
98 of 100 · LEGENDARY · WITNESSED
EDITION
E1 · witnessed 2026-10-11
MACHINE
scripts/deepfakecheck.py · 3 commands · selftest 8/8
RUNS
On your own computer. Standard library only, offline; no account, no key, no telemetry.
HOSTS
Claude · Claude Code · ChatGPT · Gemini · any AI that reads a file (HOSTS.md names each door)
LICENCE
One person, any number of their own AIs and computers · editions replace files · nothing renews · freeze, not loss
PUNCH
A5D242045764F08B82ADF7CD · sha-256 of SKILL.md · check any file against it →

CITE THIS CARDTHE DEEPFAKE CHECK CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=deepfake-check-card. Punch A5D242045764F08B82ADF7CD. Retrieved 2026-10-11.

What it is.

Images, audio and video checked for their content record before they are trusted. inspect reads inside one file for content credentials (a C2PA store in JUMBF boxes: JPEG APP11 segments, the PNG caBX chunk, the MP4 or MOV uuid box, the WAV, AVI or WebP C2PA chunk) and for XMP provenance fields, and says what is there (the manifests, the claim and the tool that made it, the signature, the assertions and actions, the digital source type in plain words) and what is missing; verdict gives one line a file: has a record, no record, or cannot tell; folder does a whole folder and counts them. It never claims to detect fakes from pixels or sound, and says so on every run; it does not check a record's signature. It waits on content records being common: most files carry none today, so "no record" proves nothing.

It carries a machine that shows its working: inspect — what is there, what is missing, the verdict; verdict — one line a file; folder — a folder, counted by verdict — 8 checks in its selftest, every command on the record. It will never say a file is fake or real from its pixels; never call a record found a record verified; never read "no record" as proof of a fake.

Six laws, written into the card.

They are in the card itself, so the AI follows them.

THE RECORD, NEVER THE PIXELSIt reads the file's record and never judges pixels or sound, so it never says a file is fake or real.
THREE PLAIN VERDICTSHas a record, no record, or cannot tell; nothing in between.
FOUND IS NOT VERIFIEDIt reads the record's parts but does not check its signature; a record found is a record to verify.
UNSIGNED WORDS, SAID SOXMP fields alone are reported as unsigned words anyone can write, change or remove.
NO RECORD PROVES NOTHINGMost files carry none today; a stripped record is named when the file still points to it.
DAMAGED IS CANNOT TELLA file cut short, failing its own checks, or of a kind it does not read is never guessed.

What rides in the card.

In the card’s own voice, with a machine that shows its working.

THE GUIDESreferences/HORIZON-TWO.md · THE-CONTENT-RECORD.md
THE MACHINEdeepfakecheck.py · three commandsinspect — what is there, what is missing, the verdict; verdict — one line a file; folder — a folder, counted by verdict. Every command prints its working and refuses a bad input; 8 checks in its selftest, every command on the record.

What it does once your AI has it.

  • inspect What is there, what is missing, the verdict
  • verdict One line a file
  • folder A folder, counted by verdict
  • On the record A JPEG carrying content credentials has a record: one manifest, its claim made with test-maker 1.0, a signature (not checked), two assertions, and a source type that says it was made by a trained model; a PNG with XMP fields only has a record of unsigned words; an MP4 record is missing its claim signature. Across eight files: 4 have a record, 2 have none (one with its stripped record named), and 2 cannot tell (a JPEG cut short and an MP3 the check does not read); no verdict ever says fake or real.

What it will never do.

It will never say a file is fake or real from its pixels; never call a record found a record verified; never read "no record" as proof of a fake.

It waits on content records being common; most files carry none today, and many platforms strip them on upload. It reads the record as the C2PA specification lays it out and does not check the signature, the certificate or the hashes that bind the record to the pixels; a full validator does that. It never detects fakes from pixels or sound. 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: who it is, its laws, when it says no, what it carries · SKILL.md — the door your AI reads first · START-HERE.md — the first five minutes, for you
  • HORIZON-TWO.md + THE-CONTENT-RECORD.md — the guides in references/, in the card's own voice
  • scripts/deepfakecheck.py — the machine: inspect · verdict · folder, standard library only, selftest 8/8
  • WITNESS.md + SELFTEST.md — every command on the record, as it printed · MANIFEST.md, every file hashed · AUDIT.md, the twelve checks
  • MY-LEDGER.md — the card's memory, on your own machine · TIN.md · LICENSE-NOTE.md · THE-CLAUSES.md · HOSTS.md
  • card.svg + card-machine.svg + card.json · card.html, the card as one offline file · the doors for ChatGPT, Gemini, Cursor, Claude Code and MCP hosts (rules/, system/, mcp/)

PROOF · SELFTEST 8/8 · 3 COMMANDS ON THE RECORD · WITNESSED IN CLAUDE 2026-10-11 · EDITION 1 · PUNCH A5D242045764F08B82ADF7CD

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 1174 PROPOSED — 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 · NOTHING RENEWS ITSELF

OFFLINE · ONE LICENCE · PUNCH-CHECKABLE · THE STANDARD

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.

One email when it is on sale, from the studio; nothing else, ever.

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 WITNESSED RUN OF 2026-10-11

deepfakecheck.py, every command showing its working.

A small Python program, standard library only. Your AI runs it; you can too. Every block below is pasted from the card’s own WITNESS.md, as printed.

inspectwhat is there, what is missing, the verdict
verdictone line a file
foldera folder, counted by verdict
selftest8 checks that prove the machine; it must print 8/8
ight small test files made for the demo by the machine's own sample builders: records shaped like the standard
(eight small test files made for the demo by the machine's own sample builders: records shaped like the standard's, signed by nothing)
inspect · WHAT IS THERE, WHAT IS MISSING, THE VERDICT
$ python3 scripts/deepfakecheck.py inspect media/with-record.jpg
with-record.jpg · JPEG · 612 bytes
verdict: has a record
  what      content credentials (C2PA), inside the file
  manifests 1 · the active one: urn:uuid:00000000-0000-4000-8000-000000000001
  claim     present · made with test-maker 1.0
  signature present (not checked: this card does not verify signatures)
  assertions c2pa.actions, c2pa.hash.data
  actions   c2pa.created
  the file says its source type is trainedAlgorithmicMedia: made by a trained model (generative AI)
  missing   nothing the check looks for
it reads the record, never the pixels or the sound: it cannot tell whether a file is faked, and it does not check the record's signature
inspect · WHAT IS THERE, WHAT IS MISSING, THE VERDICT
$ python3 scripts/deepfakecheck.py inspect media/xmp-only.png
xmp-only.png · PNG · 358 bytes
verdict: has a record
  what      XMP provenance fields only: unsigned words anyone can write, change or remove
  the file says its source type is digitalCapture: captured by a camera or a recorder
  the file says its XMP history lists 2 steps: created, saved
  missing   content credentials (a signed record)
  note      XMP names the tool it was saved with: test-maker 1.0
it reads the record, never the pixels or the sound: it cannot tell whether a file is faked, and it does not check the record's signature
inspect · WHAT IS THERE, WHAT IS MISSING, THE VERDICT
$ python3 scripts/deepfakecheck.py inspect media/clip.mp4
clip.mp4 · MP4/MOV (ISO media) · 554 bytes
verdict: has a record
  what      content credentials (C2PA), inside the file
  manifests 1 · the active one: urn:uuid:00000000-0000-4000-8000-000000000001
  claim     present · made with test-maker 1.0
  signature MISSING
  assertions c2pa.actions, c2pa.hash.data
  actions   c2pa.created
  the file says its source type is trainedAlgorithmicMedia: made by a trained model (generative AI)
  missing   the claim signature
it reads the record, never the pixels or the sound: it cannot tell whether a file is faked, and it does not check the record's signature
verdict · ONE LINE A FILE
$ python3 scripts/deepfakecheck.py verdict media/plain.jpg media/stripped.jpg media/cut.jpg media/missing.jpg
no record     · plain.jpg · no content credentials and no XMP provenance fields
no record     · stripped.jpg · the XMP points to a record inside this file, and none is there
cannot tell   · cut.jpg · the file is damaged or cut short (a segment runs past the end of the file)
cannot tell   · missing.jpg · no such file
it reads the record, never the pixels or the sound: it cannot tell whether a file is faked, and it does not check the record's signature
folder · A FOLDER, COUNTED BY VERDICT
$ python3 scripts/deepfakecheck.py folder media --report records.md
has a record  · clip.mp4 · content credentials (C2PA), inside the file
cannot tell   · cut.jpg · the file is damaged or cut short (a segment runs past the end of the file)
no record     · plain.jpg · no content credentials and no XMP provenance fields
cannot tell   · song.mp3 · not a kind of file the check reads by its structure (JPEG, PNG, MP4/MOV, WAV/AVI/WebP), and no content-credentials mark in its bytes
no record     · stripped.jpg · the XMP points to a record inside this file, and none is there
has a record  · voice.wav · content credentials (C2PA), inside the file
has a record  · with-record.jpg · content credentials (C2PA), inside the file
has a record  · xmp-only.png · XMP provenance fields only: unsigned words anyone can write, change or remove
8 files · 4 with a record · 2 with no record · 2 cannot tell
it reads the record, never the pixels or the sound: it cannot tell whether a file is faked, and it does not check the record's signature
report written to records.md

THE NUMBERS, NAMES AND DATES IN THESE RUNS ARE THE STUDIO’S OWN EXAMPLES, INVENTED FOR THE RECORD; NO REAL COMPANY, PERSON OR PROPERTY IS NAMED.

QUESTIONS AT THE TABLE

Asked before the wire.

What does it need from me?

Inspect · the record, inside the file. It asks first for the image, sound or video files (JPEG, PNG, MP4 or MOV, WAV, AVI or WebP), or the folder they are in.

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 self-test (8/8) and every command run on real example inputs, pasted into WITNESS.md as they printed; 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. It waits on content records being common; most files carry none today, and many platforms strip them on upload. It reads the record as the C2PA specification lays it out and does not check the signature, the certificate or the hashes that bind the record to the pixels; a full validator does that. It never detects fakes from pixels or sound.

What will it never do?

It will never say a file is fake or real from its pixels; never call a record found a record verified; never read "no record" as proof of a fake.

← THE INSTRUMENTS · THE FLOOR →

AT YOUR ELBOW, NEVER THE BOSS · NOT LEGAL, MEDICAL, TAX OR FINANCIAL ADVICE · NOT A PAYMENT CARD · © 2026 LUND STUDIO LLC

PAY BYAmerican Express Apple Pay Diners ClubDiscoverGoogle Pay Mastercard PayPalShop Pay Visa
CART · 0