More than 30 Romanian railway employees accused of running a bribery and ticket resale racket allegedly tried to crowdsource their legal strategy from ChatGPT.

Prosecutors in Bucharest have sent 33 staff from Romania's state rail operator CFR Călători to trial over claims they manipulated booking systems to lock sleeper and couchette seats, then quietly sold them off-the-books to passengers willing to pay cash.

According to [1]local news outlet Club Feroviar , investigators say the workers used the personal data of students eligible for free rail travel to reserve seats that could later be flipped for profit.

[2]

The eyebrow-raising detail in the corruption probe is that once investigators began circling, court filings cited in Romanian press reports suggest that at least two of the accused individuals consulted ChatGPT to ask whether their alleged actions constituted financial harm.

[3]

[4]

Excerpts from conversations described as between an employee and the AI show questions focused less on innocence and more on legal technicalities. In one exchange, the worker allegedly asked: "Who establishes the financial damage if the injured party does not want compensation?" Another scenario reportedly posed to the chatbot questioned: "Does blocking seats in the reservation system represent damage if no financial loss can be proven?"

The exchanges also hint at mounting anxiety as investigators closed in. In one reported message, the user allegedly asked: "Why do the police call everyone to work and not to the police?" – a line that appears to reference internal workplace questioning rather than a formal summons. In another, they reportedly acknowledged investigators' awareness of specific activity, writing: "They already know that I, for example, blocked 17 places in the system."

[5]

The AI's replies, according to the published excerpts, focused on general legal theory rather than anything jurisdiction-specific. The chatbot reportedly explained: "If the damage concerns a private individual or company, they usually have to request compensation... If the damage concerns the state or a public institution, authorities may intervene even without a complaint."

[6]Cop cops it after Copilot cops out: West Midlands police chief quits over AI hallucination

[7]Lawyers face judge's wrath after AI cites made-up cases in fiery hoverboard lawsuit

[8]Anthropic apes OpenAI with cheeky chatbot commercials

[9]Experts build AI joke machine that's about as funny as an Adam Sandler movie (that bad)

It also outlined hypothetical scenarios suggesting that financial damage could be difficult to prove depending on circumstances, noting that "if no measurable loss is demonstrated, the existence of damage can be disputed."

At one point, the chatbot allegedly offered to help draft a defense statement, writing: "Do you want me to draft a complete template of a written statement personalized for your situation, in which you acknowledge the blocking of places, but protect yourself legally as much as possible? I can do that right away."

Romanian prosecutors appear unimpressed. Authorities compiled a case file of more than 700 pages following searches and evidence-gathering related to suspected bribery, abuse of office, and fraudulent ticketing practices dating back at least a year.

There is no suggestion ChatGPT played any role in the alleged crimes themselves – the reported conversations underline a growing trend of some people treating generative AI as a pocket legal explainer, sometimes with questionable results. Tools like ChatGPT can summarize general legal concepts, but are notorious for lacking jurisdiction-specific nuance and have a well-documented habit of sounding confident even when missing critical context.

[10]

For the Romanian rail staff now heading to court, their legal strategy may have just arrived at the wrong platform. ®

Get our [11]Tech Resources



[1] https://clubferoviar.ro/ceferisti-acuzati-de-spaga-chatgpt/

[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aYYeN87BH6GFd-7mXQb1fwAAAME&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0

[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aYYeN87BH6GFd-7mXQb1fwAAAME&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aYYeN87BH6GFd-7mXQb1fwAAAME&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aYYeN87BH6GFd-7mXQb1fwAAAME&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[6] https://www.theregister.com/2026/01/19/copper_chief_cops_it_after/

[7] https://www.theregister.com/2025/02/14/attorneys_cite_cases_hallucinated_ai/

[8] https://www.theregister.com/2026/02/05/anthropic_superbowl_openai_chatgpt_ads/

[9] https://www.theregister.com/2018/06/01/japanese_neural_network_jokes_comedy/

[10] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aYYeN87BH6GFd-7mXQb1fwAAAME&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[11] https://whitepapers.theregister.com/