AI is flooding the EU Parliament: sending is free, checking is not

An Irish MEP issues an internal policy banning AI for votes, plenary speeches, press releases and questions to the Commission. The document announcing the ban? According to heise, it discloses that it was drafted with AI help. Sounds like satire. It isn't. To me it's the most honest moment of the whole debate: the problem isn't AI. The problem is that nobody drew the line where it stops.

What happened

heise reported on November 13, citing the Brussels outlet Rapporteur, that the European Parliament can barely keep up with AI-generated submissions. Secretary-General Alessandro Chiocchetti warned of a rapid rise in amendments, parliamentary questions, petitions and requests for access to documents, and attributed it to generative AI. Complaints to the European Ombudsman are up 50 percent because of AI, according to an unnamed senior EU official, and many of them are poorly written. Interpreters found references to laws that don't exist in MEPs' texts. According to The Next Web, Parliament's lawyers needed two days to disprove a piece of wrong legal advice from an AI tool. President Roberta Metsola had already warned in early November that an AI-drafted amendment is no substitute for real debate. On November 12, Parliament adopted a resolution by 405 votes to 133, with 60 abstentions: AI in the institutions only under human oversight, never purely automated political decisions.

Effort was the filter. It's gone.

For decades, one thing quietly did the sorting: writing took time. If someone filed a complaint, an amendment or a petition, they usually meant it. Volume was a signal. Generative AI removes that cost. But only for the sender. Reviewing costs exactly what it did before. Two days of legal time for one invented answer, that's the bill.

Brussels feels it first because a public body has to process every submission. Companies have the same asymmetry, just quieter: contact forms, job applications, complaints, tenders, support tickets. If your process relied on effort to separate signal from noise, AI is now showing you the cracks in the foundation. It widens them, it doesn't fix them. Translation: a spike in inbound leads on your dashboard is no longer good news until someone has checked the quality.

Draw the line at the signature, not the tool

What I find interesting is where the MEPs themselves draw the line. Sweden's Jörgen Warborn uses AI for research and calls it a productivity gain. Barry Andrews bans it wherever his name and his vote are attached. And Chiocchetti explicitly wants AI to help sort thousands of amendments. In my view, that's exactly the right rule of thumb: AI can sort, summarize and prepare. Judgment, accountability and final sign-off stay with a human. Parliament's upcoming guidelines, per The Next Web, are refreshingly plain: disclose AI use, never put sensitive files into public tools, check everything before it goes out.

The catch: when AI writes on both sides, nobody is talking anymore. A parliamentary assistant quoted by heise warns of largely human-free communication between institutions. Inside a company it looks like this: the AI-written complaint gets an AI-written reply, and the one real, justified complaint drowns in the noise.

What I'd do now

• Count your inbound channels: where has volume jumped unusually since the start of the year? Forms, applications, complaints. Spot-check quality first, celebrate later.
• Triage before you reply: let AI sort, summarize and flag red flags like invented legal clauses or contract numbers. The decision and the signed reply stay with a person.
• Steal the Brussels rule: three sentences in your team handbook. Disclose, no sensitive data in public tools, check everything before it ships.

Bottom lineWhen writing costs nothing, attention becomes the scarce resource, and whoever doesn't organize it drowns politely.