Law firm flags court rulings citing fake Supreme Court precedents

Rulings in 24 hours cite non-existent laws and fake Supreme Court precedents

Can a judge issue a verdict on the day of trial without reviewing medical reports or hearing witnesses? A family-run firm handling civil, labor, incivil, and divorce cases argues this is happening: resolutions against the administration arriving in less than 24 hours, citing non-existent legal articles and Supreme Court judgments with official-looking case numbers and dates that do not appear in CENDOJ (the Spanish Judicial Data Center). The core accusation is that AI has entered the judiciary without supervision and no one is cleaning up the trail.

The account is detailed. The PDFs of these rulings contain bold asterisks, code blocks, and copy-paste symbols that reveal the text's origin. They say the judge was not present in the courtroom; the resolution is drafted from prior written submissions, so if a witness confesses something new or a document is presented at trial, the judgment ignores it. Hence the constant requests for nullity and the extra cost for clients, who risk their pension or home while someone signs off hastily.

What the CGPJ says and what judges risk

The most common institutional rebuttal is that this is why the General Council of the Judiciary (CGPJ) exists, and a judge signing a ruling with invented precedents risks their position. This is not a minor argument. The judicial career has a disciplinary regime and perversion of justice is a crime, so the question is not whether it is possible, but whether the system detects and sanctions it. Some recall that only a handful of high-profile judges have been removed from service in recent decades, fueling a sense of impunity.

The counterargument is that AI did not invent anything: copying and pasting jurisprudential filler and padding arguments existed before ChatGPT. What is new is the estimulante ilegal and the digital footprint. Before, bad arguments were invented; now, precedents are invented. The difference is that an invented precedent can be verified in two minutes, whereas a bad argument cannot.

The lawyer winning cases with databases vs. the one using chatbots

The conversation shifts to an uncomfortable topic: it is not just judges. Some lawyers admit to using two legal databases as they always have, while others, according to colleagues, rely on chatbots without verification. Cross-accusations between public defender lawyers and private practitioners are classic: some say the state-funded duty is poorly paid, leading to lack of time; others respond that the price does not justify shoddy work. In the middle lies the suspicion that the problem is not the tool but the lack of oversight.

The most striking detail involves hidden characters. Those reviewing digital documents find formatting symbols in rulings that should not be there. It is proof that the text is pasted without cleaning. A judge can make mistakes; what is untenable is that the error travels with the copy-paste mark.

Is this widespread or isolated?

There are no official figures in the material. What exists are first-hand testimonies, some very specific, and a clear division between those who see it as widespread practice and those who consider it unlikely. Digital Justice 2.0 already allows signing with a click; the question is how many clicks happen without reading. Until there is a count of rulings annulled due to non-existent precedents, the discussion remains in the realm of founded suspicion.

The final irony comes from the system itself: if AI hallucinates laws and the judge signs without reading, the only person working in the chain is the appealing lawyer. And that lawyer gets paid.