A journalist using artificial intelligence (AI) to generate an article is like a chef in a restaurant serving you a reheated ready meal from a supermarket. It’s fundamentally phoney. This is true even if the contents of the article happen to be accurate – because the journalist doesn’t really know whether they are true or not.

Anyone I know who has worked with Peter Vandermeersch has a very high opinion of him. From 2019 until last year he was one of the most influential people in Irish media, first as publisher and then as chief executive of Mediahuis Ireland. The Belgian company publishes the Irish Independent, the Sunday Independent, the Belfast Telegraph, the Sunday World, Sunday Life, The Herald and 11 regional newspapers.

Vandermeersch’s job was to make those titles fit to survive in the digital era. Anyone who cares about democracy and the diversity of information and opinion that keeps it alive would recognise that this is more than a business imperative. It’s a political, social and cultural obligation.

Vandermeersch is no cub reporter. He’s 65. He was a foreign correspondent in Paris and New York before serving as editor in chief of the Brussels-based newspaper De Standaard in 1999-2010 and then of Amsterdam-based daily NRC Handelsblad in 2010-2019. He’s a sophisticated, cultured and thoughtful man who knows pretty much everything there is to know about journalism. Which makes what happened to Vandermeersch last week rather hair-raising. He was suspended from the position he was given after he stepped down from the Irish operation last year as “Journalism and Society” fellow – essentially Mediahuis’s in-house intellectual.

Price of diesel and petrol set to fall from midnight as Government agrees new excise duty measures

‘I want the mess gone’: Couple in Co Meath demolition case swear not to interfere with works

How YouTuber Stephen McCullagh tried to get away with the murder of his pregnant partner Natalie McNally

‘I don’t envisage lying in front of bulldozers’: Man wants reprieve for €300,000 Sutton extension

His new job was to think about the big existential questions. As he explained when the appointment was announced by Mediahuis last August: “I want to further strengthen Mediahuis’s position as a thought leader, addressing some of the industry’s key challenges – such as the relevance of journalism, its role in an era of polarisation, trust, press freedom, engaging with diverse audiences and ensuring a sustainable future for local journalism – while exploring the responsible use of AI in newsrooms.”

Part of Vandermeersch’s role was to publish regular blogs under the Press and Democracy rubric reflecting on these topics. In one of the first blogs he concluded, with regard to the threat posed by the use of large language models to generate content, that: “The battle lines are drawn. What’s at stake is not just the future of the media, but the future of truth itself.”

Last week NRC, the Dutch newspaper he used to edit, published an analysis of Vandermeersch’s blogs. It found that “in 15 of the 53 blog posts he wrote, there are quotes that cannot be found in the publications from which Vandermeersch claims to have obtained them, such as news articles and scientific studies. Seven of the quoted individuals confirm that they did not make the statements attributed to them in these publications – nor, as far as they can remember, elsewhere. Altogether, this involves dozens of quotes.”

The irony is that one of the recurring themes in these very articles is the absolute requirement for journalists to verify what they write. Responding to a report on the collapse of trust in journalism among young people, Vandermeersch set down a basic rule: “Show the process. Let young people see how stories are verified.”

[ AI’s war on reality: what now when you can’t even trust your own eyes?Opens in new window ]

On the threat from AI to the economic viability of journalism, he wrote that “a society that allows its news to be mined and regurgitated without remuneration doesn’t merely impoverish journalists; it starves itself of truth. Freedom of expression is meaningless if there are no reporters left to verify the facts.”

Yet, as he wrote on the blog after his suspension, “I used AI language models such as ChatGPT, Perplexity and Google Notebook while writing. I was enthusiastic about the possibilities these tools offered and wanted to experiment with them extensively. Even I – with all my years of experience and knowledge – fell into the trap of hallucinations. I summarised reports using AI tools and worked from those summaries, trusting they were accurate. In doing so, I wrongly put words into people’s mouths ...”

A journalist using “quotes” regurgitated by machines to write about the dangers of AI is like a doctor treating patients by injecting them with a deadly virus. Vandermeersch actually wrote about fake quotes from made-up individuals appearing in British newspapers: “Their quotes cropped up in articles on energy prices, transport, home renovation, health and consumer behaviour. Only on closer inspection did it become clear that they did not exist.”

[ The inconvenient truth about artificial intelligenceOpens in new window ]

Attributing fake quotes to real people creates a double harm: the non-existent quotes are released into public discourse and the individuals who are supposedly being quoted are potentially traduced. One of AI’s talents is to ascribe stupid or disreputable statements to people who did not make them, and then feed them back into its own endless loops of lies. AI is not getting more accurate – it’s just getting better at making false quotes and “facts” look plausible.

Here’s a quote that isn’t fake: “A newsroom that wants to retain credibility must invest in time, verification and scepticism. Without that, journalism loses its role as a countervailing force – and becomes a conduit.” That’s also from Vandermeersch and it is 100 per cent accurate.

So what can have come over him? I suspect the answer lies in his life as a newspaper business executive who can’t help having an eye on the bottom line. That part of his self-described role – “exploring the responsible use of AI in newsrooms” – is not an abstract intellectual quest. It’s a search for ways to cut more jobs and make higher profits.

The responsible use of AI in newsrooms is already clear: don’t do it. Verification and scepticism are what journalists live by and what machines are incapable of. As someone really did say, what’s at stake is the future of truth.

[ Peter Vandermeersch controversy illustrates AI challenge more vividly than anything he wroteOpens in new window ]

© The Irish Times