The EU, AI and Journalism
The AI Act came into force on 1 August 2024. The EFJ participated to the plenary dedicated to drafting the Code of Practice on Generative AI for the EU AI Office, alongside nearly 1,000 other organisations and individuals. The plenary will set guidelines for the responsible use of Generative AI within the EU’s AI regulatory framework.
EFJ members have also discussed in expert groups the impact of generative AI and have raised concerns on controlling and regulating the use of AI in newsrooms. The EFJ Authors’ Rights Expert Group (AREG) drafted recommendations on the right of identification and personality, fair remuneration and licensing.
AI is a useful tool for journalists. It can be used to transcribe long interviews, sift through large datasets, allow local media outlets to generate visuals if they do not have their own photojournalists, and more. Just as all digital tools must be used with caution, so does AI – and perhaps it requires even more diligence.
“Some experts say AI is just a hype, others say it will define the competitiveness of Europe’s future. In journalism, while generative AI will definitely be a rapidly evolving tool, which must be accompanied by training and ethical guidelines, it may support independent journalism. But only if regulated and in respect of copyright and fair remuneration! Otherwise, it may destroy the fundamental business model for journalism”, said EFJ Director Renate Schroeder.
The new EU Commission, under the leadership of President von der Leyen, intends to focus on Europe’s competitiveness with the US and China, pushing for European AI models that aim to meet the citizens’ needs. EU Commission President has said “Europe must also become a global leader in AI innovation,” and “Europe is already leading the way with the EU AI Act, ensuring AI is safer and more trustworthy”.
This next EU legislature will be about AI and EU’s competitiveness. 17 Commissioners with technology in their profiles reflect this focus. The European Parliament’s Committee on the Internal Market and Consumer Protection (IMCO) AI Working Group also shows the Parliament’s eagerness to get ahead.
AI models have had free-range on collecting copyrighted text and data. They use these materials to recreate their outputs, violating copyright standards set out in the EU Copyright Directive and national law. Moreover, they do not at all fairly remunerate the creators including journalists of this data from which AI developers make their profits.
The Copyright Initiative published a study which showed that the current training of AI models is copyright infringement. For the first time, a legal scholar and a professor of AI worked together to shed light on the so-called black box and prove that the common applications of generative AI on the market are not compatible with current law.
“We now have it in writing that the practice of AI training by the big tech companies is simply illegal. A ‘Green’ AI must reflect our social values and norms and avoid violations of the law in advance.” – DJV Federal Chairman Mika Beuster
Some publishers are creating licensing agreements (including News Corp, the Associated Press, Axel Springer, the Financial Times, Dotdash Meredith, The Atlantic and Vox Media). Others (including the New York Times, The Intercept, and eight newspapers owned by Alden Global Capital) are suing AI developers. Those who create licensing agreements say that this gives them control of how the news is handled. Those who sue claim to do it for fair journalism. What is certain is that the legal route is long, and copyright laws in the EU and US complicated.
This is extremely worrying for journalism and journalists, as the future of the already fragile business model of independent media is further jeopardised by both violation of copyright and fair remuneration and the use of generative AI possibly replacing journalistic paid content by offering free open data tailor made for all.
The recent 2024 Media Freedom Rapid Response (MFRR) report highlighted spoofing as a main concern for journalists during the first six months of this year. Spoofing attacks involve altered photo, voice, or video material including Artificial Intelligence (AI) generated deep fakes, cybersquatting, as well as spoofing websites and articles that pose to be created by legitimate news outlets or media workers. Spoofing attacks are often used to discredit journalists and media outlets, to spread disinformation, as well as to promote commercial products with fraudulent advertising.
“Spoofing increasingly affects the credibility and integrity of the media, but also threatens the fundamental right to access information free from manipulation. Spoofing is not new, but it is a growing phenomenon with the rise of AI tools, and platforms are doing too little to remove spoofed content, especially when it is reported. Identifying and prosecuting perpetrators of spoofing is also crucial to ensure a safe environment for media professionals and reliable information.” – Camille Magnissalis, EFJ Press Freedom Monitoring Officer
Although AI has been used in its most ‘basic’ form in newsrooms for several years (capturing sports results, elections, etc.), more and more European media are interested in ‘deep learning’ versions of AI, with the aim of disseminating articles and audio and video reports at lower cost.
The EFJ was part of the Paris Charter on AI, led by Maria Ressa and organised by RSF. Various European media councils have set standards on the use of AI in journalism. The rules and know-how are there, what’s needed now is training, enforcement and adaptation to local and other news organisations throughout Europe.
