Irish Writers Union and European Writers’ Council warn that tech companies are failing to meet even basic EU transparency requirements, putting writers’ livelihoods, Ireland’s creative industries and the wider public at risk.
Dr Conor Kostick, representing the Irish Writers Union and, as our Board Member, the European Writers’ Council, took part at a hearing at the Oireachtas Joint Committee on Artificial Intelligence on Tuesday, 17 November. Two of the main messages: AI companies are failing to meet even the minimum transparency requirements set out under European legislation – and the need for respecting the EWC’s ART-Principle.
In dialogue with Writing.ie Dr Kostick said: “We are witnessing a fundamental threat to the rights of writers and creators, but this is not simply an issue for the literary community. It is an issue for society as a whole.
“AI companies are building enormously valuable technologies using creative works produced by authors, often without their permission or payment. At the same time, serious questions remain about whether these companies are meeting even the most basic transparency obligations required under European law.
“AI companies cannot be allowed to operate above the law. Innovation should not come at the expense of creators’ rights, public accountability or the creative industries that contribute so much to our economy and culture.”
Dr Kostick urged the Oireachtas Committee to adopt the European Writers’ Council’s three core principles for the regulation of AI, known as ART: Authorisation, Remuneration and Transparency.
Read the full intervention – or watch the recorded hearing here.
Authorisation: the consent of the writer is needed before our books are used in AI training. Authors must retain the freedom to say “no” and remuneration after the fact is no substitute for the that freedom.
“Authors must retain the freedom to say no. Offering payment after the fact is no substitute for taking away that fundamental right. Creative work belongs to its creator, and technology companies should not be entitled to use it without permission,”
Remuneration: Authors are the source of a 24.5-billion-euro book economy in Europe. Our intellectual labour and the wellspring of this sector of the economy is being dismantled without remuneration.
“We cannot allow an industry built on human imagination, creativity and intellectual effort to be dismantled by technologies that have benefited from that very same work.”
Transparency. Just as it is in the public interest that food is properly labelled, so that we know what we are putting into our bodies, so we need to know what we are putting into our minds. Everything created by generative AI should carry that information and – like with food – the labelling should be detailed.
Why would I make these my priorities? Because 2026 has seen a tsunami of AI outputs rush over the creative sector and 2027 will see an exponential growth in the publication of AI simulating books, music, and films. We are going to be absolutely submerged in the stuff. This is rapidly undermining author incomes, eliminating many creative roles like translators, book cover creators, and no doubt a range of other artists too. To give the case study that I’m most familiar with: a recent paper has shown a 17 per cent decline in income going to human-written books from 2023 – 2025 on Amazon’s Kindle Unlimited platform (https://arxiv.org/pdf/2607.20349).
Irrespective of the question of whether AI products are any good, readers are seeking human-to-human interaction. As one person put it in on online forum: I literally feel like crying. I don’t know why. This is the first time it happened to me, and I now know why readers are always complaining about finding an AI written book. I have never felt so betrayed in my life. Never.
Detailed labelling of AI products would allow us all to find the human-created artistic work that we desire. It would also help the public cope with political misinformation that can so easily and persuasively be manufactured by AI.
Transparency as to the process by which AI products are made is also an essential perquisite for artists – in our case authors – to obtain remedies for the theft of our copyrighted works in the training of these models. It is well established by the disclosures in US courts (for example during the Anthropic case) that many, if not all, of the large language models in use today were trained on copyrighted works.
Clear acknowledgement of the materials on which these models were trained will allow authors to seek redress for the theft of our books, musicians for the theft of their music, film makers for the theft of their films.
When I use the language of theft, by the way, the theft I’m talking about is not just from the individual copyright holder. It is from the public, the taxpayer. The EU has invested a substantial amount of taxpayer’s money in the arts and in Ireland we have schemes like the Basic Income for the Artist as well as Arts Council bursaries and other awards that have been vital in the creation of our art. When foreign big tech companies help themselves to this art without any attempt to licence the right to use it, they are stealing publicly funded art worth billions for their companies.
If companies are failing to meet even relatively modest transparency requirements, we have to ask how seriously they are taking their wider legal and ethical responsibilities. Legislation is only meaningful if it is enforced. We need European and Irish authorities to ensure that AI developers are held accountable, rather than allowing compliance to become little more than a box-ticking exercise.”
I appreciate that this committee has already stated that The EU Copyright Directive needs to be strengthened. I would add that the recent Advocate General’s Opinion in Like Company vs Google Ireland (c-250/25), indicates continued uncertainty about the misuse of the exception in the Copyright Directive and we would urge implementation of this aspiration.
The committee was also right to say: Ireland must not shy away from the EU AI Act or try to dilute it. We should treat it as a minimum baseline for national AI regulation, not a maximum standard.
What the committee might not be aware of is that a recent study by researchers in TCD (led by Harshvaardhan Pandit, TCD AI Accountability Lab) concurs with the committee’s conclusion, having looked at compliance with the AI Act and found that there were over 30 AI models, the latest ones, which had not provided submissions for that Acts low-bar transparency requirements in the so-called “template”. Most of those that had made submissions were, and I quote, “bad faith implementation, a wilful exclusion of the requirement.”
In other words, even a minimum regulation of AI tech companies is not being enforced.
We ask the Committee to recommend that Ireland:
- Advocate at EU level for effective authorisation rights in the application of the TDM exception to generative AI.
- Support stronger transparency requirements covering both AI training inputs and AI-generated outputs.
- Ensure that authors’ rights and remuneration are properly addressed in any AI licensing framework.
- Examine mechanisms for affordable and effective enforcement, recognising that individual writers cannot realistically be expected to litigate against major technology companies.
- Use its position as the European base for significant technology companies to support meaningful compliance with EU copyright law.
.
Relevant resources


