2026-11-20 · ← News
Three AI disputes: fake IDs, an agent suing Meta, and a detector in court
A 404 Media podcast links OnlyFake, a lawsuit produced by a Muse agent, and the US Department of Justice's use of the Pangram detector. These are not one grand scandal, but three concrete points where cheap automation collides with institutions built on trust.
The image could not be loaded.
A 404 Media podcast brings together three separate reports: the creator of OnlyFake, a small-business owner's lawsuit against Meta prepared by his Muse agent, and a dispute among US lawyers over the Pangram AI-text detector. The shared theme is narrow and practical: tools made document production cheap faster than institutions changed how they verify those documents.
OnlyFake turned ID forgery from a craft into a 15-dollar service
Yuriy Nazarenko told 404 Media that he first edited digital passport copies by hand and later automated the process. OnlyFake could generate images of nonexistent documents placed on a desk or carpet, making them look like phone photographs designed to pass identity checks.
According to a US Department of Justice announcement cited by the report, the service made more than 1 million dollars and charged 15 dollars per fake. Nazarenko pleaded guilty and, after a year in prison, also described use by fraudulent North Korean IT workers. AI did not create a new category of fraud here. It removed the skill and price barriers.
A Muse agent turned a suspended page into a 75,000-dollar lawsuit
Business owner Joe Thomas gave a Muse agent access to his company's Facebook pages. Meta suspended the accounts after treating the agent's activity as a possible attack. Thomas says he lost at least 25,000 dollars in work during the outage. The lawsuit filed in South Carolina seeks 75,000 dollars.
The agent, called Baby Mia, allegedly prepared the complaint, contacted politicians, agencies, and journalists, and sent legal demands. A real lawsuit does not mean it will succeed or that an agent has legal personhood. It shows how one user can flood several institutions with documents and requests when automation has no natural stopping point.
Pangram entered the courtroom as one party's allegation
In the third report, 404 Media says US Department of Justice lawyers used Pangram while arguing that opposing counsel generated filings with AI. Government lawyers called the approach “defense by ChatGPT” and described filings as “AI slop.” Those are one party's claims in court documents, not a judicial finding about authorship.
A detector can flag suspicious text, but it cannot by itself decide whether a legal argument is sound, whether a lawyer broke a rule, or who wrote a particular sentence. The danger is that a dispute over facts becomes a dispute over a tool's score that the other side may be unable to reproduce independently.
Institutions must measure harm, responsibility, and evidence separately
These three stories require different responses. Fake IDs demand stronger identity verification. Agent-generated filings require a clear record of who authorized each action and bears responsibility. AI-text detection requires courts to examine concrete errors and evidence rather than rely on a probabilistic label.
The next signals will come from court rulings, electronic filing rules, and changes to KYC systems. If institutions respond with another automatic filter alone, two black boxes will race each other. One produces documents, the other labels them, and the human in the middle still cannot tell what to trust.
Lilith's verdict
A fraudster, an angry business owner, and a government lawyer can now place a stack of documents on an institution's desk with one click. Anyone answering with another automated stamp has just called a black box as a witness.