Inside The Dáil’s Secret Struggle: How AI Could Make or Break Ireland’s Political Future—And What No One’s Talking About

Is this really Gavan Reilly typing out this column—or is it some clever digital doppelgänger lurking behind the scenes? In an era where artificial intelligence buzzes beneath the surface of nearly every headline, the very notion of authorship demands a skeptical eyebrow. You see, OpenAI’s latest ChatGPT version has become a virtuoso mimic, capturing an author’s unique voice so well that even the old giveaways—those pesky em dashes—have vanished like magic. So, when you’re reading a “Gavan Reilly” column, how sure can you be it’s actually Gavan behind the words? It’s a head-scratcher that raises bigger questions about authenticity, trust, and what it means to communicate in today’s political and editorial arenas. Meanwhile, AI is quietly weaving its way into everything from political speeches to legislative drafting, promising to speed things up but also casting shadows of “hallucinated” facts and accidental misinformation. Before we all dive headlong into this brave new world powered by bots, maybe it’s time to press pause, question the rules, and figure out how to navigate AI’s fingerprints on our public discourse. Oh, and in case you’re wondering—yes, this piece really is penned by me. LEARN MORE

Firstly, a necessary reassurance: yes, this column for Business Plus has been written by Gavan Reilly.

Why is that necessary? Because technological change now means you can’t take for granted that the name on a piece is actually the name of the author.

Version 5 of OpenAI’s landmark ChatGPT software excels at replicating a given author’s own original style of writing.

No more telltale em dashes (long dashes commonly used by AI) giving it away.

Now, it seems, you can ask for a paragraph in the style of Person X and out comes a reasonable imitation.

Feed it some copies of your own writing style and, hey presto, a column.

That raises some obvious questions for the likes of your humble columnist. It may make the life of political speech-writers slightly easier.

Party leaders generally have their speeches written by a small number of trusted lieutenants who, it is said, ‘have their voice’.

Life is much easier if a robot can produce a rough draft for the ghostwriter to finesse later.

Some politicians in high office already admit to using AI as a professional aid.

Ulf Kristersson, the prime minister of Sweden, has said he used ChatGPT and a French counterpart, Le Chat, to scrutinise his own thinking.

“What have others done? Should we think the complete opposite?” he previously said.

For the record, he insisted that his prompts never included any sensitive government data.

Though not yet having an official government application here, AI is now a feature of Irish politics too.

Earlier this summer, three Sinn Féin TDs told the Dáil that the stock price of United Natural Foods (UNFI) had surged on the NYSE partly because of profiteering at its Irish subsidiary SuperValu.

Readers of this magazine will immediately spot something amiss: SuperValu is not owned by UNFI, but rather by Ireland’s own Musgrave Group.

Eagle-eyed observers noted, though, that UNFI does own a chain called Supervalu (with a small v) — and that if asked to generate a polemic about profiteering in Ireland, certain AI tools would throw up text that didn’t realise the difference.

Asked if AI was to blame, Sinn Féin cited “human error”.

If the mistake is the fault of AI, it’s not the first instance of AI input to Dáil proceedings.

In a debate this year about technology, Paul Gogarty, the independent TD, read a statement ChatGPT had written for him.

Shorn of its telltale em dashes, the official transcript doesn’t give the game away — you’d only know AI had written the speech because Gogarty himself said so.

But that may not be the first time. In my research for a book published earlier this summer, several politicians told me they suspected AI was being used more regularly to generate lazy talking points for speeches, and unwittingly laundering fake news in doing so.

For example: a rumour of alleged misbehaviour, say at an IPAS centre, is shared with a TD’s office.

That TD’s assistant asks ChatGPT (in an earlier edition) to generate a speech that mentions this incident.

The AI tool cannot distinguish if the allegation is true, but takes its instruction in good faith.

The TD reads the resultant speech into the record. What starts as a mischievous rumour is given permanent credence.

If harnessed correctly, AI could have incredible effects on Irish policymaking.

Take the time consuming act of drafting legislation. This suffers from a perpetual bottleneck of legislative draughtsmen.

AI could, if deployed well, dramatically increase the throughput of those finite drafters.

But before generative AI becomes an everyday part of political business, as it will with every workplace, it might be wise for politicians to pause the usual hostilities and consider a charter for its use in Leinster House.

Otherwise, I fear, hallucinations about supermarket ownership might not be the only time our leaders enter debates on false pretences.

I’m the real Gavan Reilly and I approve of this message.