⚡ Quick Summary
Published by the Council of Europe and adopted by its Steering Committee on Media and Information Society (CDMSI) on 30 November 2023, this resource sets out guidelines on the responsible implementation of artificial intelligence systems in journalism. It defines "artificial intelligence system", "journalistic artificial intelligence system", "artificial intelligence user", "artificial intelligence technology provider", "artificial intelligence subject" and "artificial intelligence audience", and anchors the use of AI across the journalistic value chain — research and data analysis, production, dissemination, personalisation and audience engagement — in Article 10 of the Convention for the Protection of Human Rights and Fundamental Freedoms together with Articles 8, 9 and 14.
The purpose is to carve out principles for media organisations and media professionals, and to offer guidance to AI technology providers, platforms, States and national regulatory authorities. Six chapters address the decision to implement AI, identification and acquisition, incorporation into professional and organisational practice, use in relation to users and society, responsibilities of external technology providers and platforms, and obligations of States. Annex 1 provides a procurement checklist and Annex 2 an overview of existing Council of Europe guidance.
🧩 What’s Covered
The document moves from framing to definitions, then to six chapters of guidelines, closing with two annexes.
- Introduction: frames AI as useful across the journalistic value chain, recalls the democratic role of journalism and the rights and duties under Article 10, cites Recommendation CM/Rec(2022)4, and calls for AI to be demystified.
- Definitions and scope: sets out the six definitions, notes that the AI system definition comes from the Consolidated Working Draft of the Framework Convention on AI and may change, and states that the Guidelines address the decision to use, identify and acquire journalistic AI systems rather than design and development.
- Guidelines 1 – the decision to implement: mission-driven and editorial decisions, clear accountability (typically the editor-in-chief), a systematic risk assessment, a participatory process with a "problem owner", proof-of-concept testing, and a circular review process.
- Guidelines 2 – identification and acquisition: procurement or in-house development, evaluation of data availability, data fairness and data quality, privacy and data protection compliance, measures against bias and stereotypes, and assessment of the provider's data practices.
- Guidelines 3 – incorporating AI into practice: hiring and upskilling without simply replacing trained journalists, editorial oversight of processes as well as outputs, continued risk assessment, disclosure and standardised labelling, metadata quality, ongoing training, and a pioneering role for public service media.
- Guidelines 4 – users and society: responsibilities and duties towards citizens, a transparent vision expressed in self-regulatory and organisational codes, traditional journalistic values that remain relevant, audience-centred values, accountable personalisation, value-sensitive design, and a five-element process running from documented value identification to public accountability.
- Guidelines 5 and 6 – providers, platforms and States: respect for editorial autonomy, warning time and assistance from technology providers; platform obligations drawn from CM/Rec(2022)4, CM/Rec(2022)11 and the guidance note on prioritisation of public interest content; State duties on funding diversification, provider choice, standards, standard procurement clauses, negotiation imbalances and data, media and AI literacy.
- Annexes: Annex 1 lists procurement themes and questions (training data, model, data ownership, data storage, liability, human oversight, responsible development, infrastructure, continuity, pricing, mutual support, environment); Annex 2 lists relevant Council of Europe conventions, recommendations, declarations, guidance notes and research, and tabulates AI-related concerns against their sources.
💡 Why it matters?
Media organisations deciding whether and how to adopt AI need to connect procurement choices, editorial oversight and human rights obligations. These Guidelines supply that structure: a systematic risk assessment that informs acquisition, disclosure towards audiences, oversight that cannot be replaced by code, and a checklist of contract questions on training data, liability and continuity. Responsibilities are also assigned upstream, to technology providers and platforms, and to States and (self-)regulators asked to support standard clauses, research and AI literacy. The document ties its guidance to the ECHR and to Council of Europe instruments, notably CM/Rec(2020)1 and CM/Rec(2022)4.
❓ What’s Missing
Design and development are excluded from comprehensive coverage, on the grounds that both are highly specialised and task specific; only broad aspects are addressed. The document states that not all procurement questions are equally relevant to all organisations and that the Annex 1 checklist is not exhaustive and should be a living document. The AI system definition is provisional, taken from the Consolidated Working Draft of the Framework Convention on AI and subject to later adjustment. The text is addressed to Council of Europe member States and their media sectors, and its legal references are European. Resource differences between large and small organisations are noted, but no resourcing models are offered.
👥 Best For
Best suited to editors and editorial leadership weighing the adoption of AI in a newsroom; to legal and compliance teams drafting or reviewing AI procurement contracts; to technology providers and platforms working with news organisations; and to regulators, self-regulatory bodies and national authorities asked to support standards, procurement clauses and literacy initiatives in this sector.
📄 Source Details
Guidelines on the Responsible Implementation of Artificial Intelligence Systems in Journalism, adopted by the Council of Europe's Steering Committee on Media and Information Society (CDMSI) on 30 November 2023 and published by the Council of Europe; the imprint records March 2024. The extracted PDF runs to 41 pages and contains the introduction, definitions, scope and purpose, six chapters of guidelines and two annexes; no individual authors are named. A French version, Lignes directrices sur la mise en œuvre responsable de systèmes d'intelligence artificielle dans le journalisme, is noted. The input was a text extraction of all 41 pages. No URL for the document itself is printed; the back cover shows only www.coe.int.
