A new briefing investigating AI in the CCS and assessing whether existing EU policies address AI challenges

IRIS 2026-9:1/2

Justine Radel-Cormann

European Audiovisual Observatory

The research department of the European Parliament’s Committee on Culture and Education (CULT) published a briefing on 23 September 2026 on artificial intelligence (AI) in the cultural and creative sectors (CCS).

The briefing examines the impact of AI on the CCS and the European Union’s (EU) policies in this area. It identifies gaps between the challenges posed by AI and the existing EU legislative framework, and makes recommendations to address them.

The AI Act establishes a horizontal, risk-based framework that applies across industries. It is complemented by the Directive on Copyright in the Digital Single Market, the Digital Services Act, the Data Act, the Data Governance Act and the General Data Protection Regulation (GDPR). The briefing recommends strengthening the AI Act’s transparency obligations, including by requiring comprehensive documentation of data sources.

Many of the challenges facing the audiovisual sector are shared with music and publishing. These include potential job losses and the risk that professionals may lose skills. Professionals also need access to AI training to remain competitive. One of the briefing’s high-priority recommendations is to safeguard entry-level jobs and core skills.

AI could also amplify bias, misrepresent people and cultures, and distort cultural representation. The Audiovisual Media Services Directive (AVMSD) supports cultural diversity by promoting the visibility and discoverability of works in video-on-demand catalogues. However, there is no guarantee that AI models will represent different cultures and identities, creating a risk of cultural homogenisation. The briefing recommends setting a binding cultural and linguistic diversity obligation for AI systems in the AI Act. 

There are also concerns about the use of copyrighted works to train AI systems. The briefing notes that, under the current framework, using copyrighted works for AI training is not considered copyright infringement if rights holders have not explicitly opted out of their works being included in training data. The briefing recommends extending opt-out mechanisms beyond data used for AI training to cover any data collected for exploitation, and adapting those mechanisms to reflect the sectors’ operational and licensing realities.

AI may also put actors at risk if avatars are used to replace extras or create synthetic actors, digital replicas or voices. Artists may face pressure to grant rights for the use of their work in AI training. The GDPR protects personal data, and its rules may give rise to civil liability when AI models are trained on people’s likenesses or voice recordings. However, there is no reliable way to verify whether personal data has been excluded from training data. The GDPR protects living individuals, leaving a gap in relation to deceased persons. The briefing recommends strengthening enforcement of personal-data protections and creators’ rights in AI development, and treating personal data inferred, generated or reconstructed by generative or agentic AI systems as personal data.

The briefing identifies more than 20 recommendations, ranked from low, to medium to high priority.