Who sets the standard
Ireland chairs the Council of the European Union until the end of December, with competitiveness, values and security as the three priorities of its presidency. For news organisations, those priorities invite a practical question: who defines the conditions under which AI companies can use journalism, and who checks that they respect them?
The Irish agenda combines AI adoption, copyright and the competitiveness of European businesses, while its programme also prioritises child safety and privacy-preserving age verification. Ireland intends to advance a review of the CDSM directive and seek agreement on the broader Digital Omnibus by the end of 2026. These objectives require decisions about access, evidence, responsibility and cost. Each decision affects the organisations that produce information and the people who depend on it.
On 14 October, NewsBrands Ireland and the Irish Recorded Music Association will bring copyright and human creativity into that discussion at a conference in Dublin devoted to licensing, trust and effective enforcement. News organisations have an opportunity to explain what such a market requires, from permission and attribution to reliable usage records and a sustainable price. The discussion will be more useful if the sector can describe how those requirements work in practice.
This edition considers those requirements alongside newsroom interoperability, AI discovery, media intelligence and synthetic influence. My working proposition is that standards deserve the same executive attention as contracts: they determine what a buyer can request, what a supplier can prove and how the parties resolve a dispute. The sector’s negotiating position depends partly on its ability to specify its own terms.
European Commission President Ursula von der Leyen and Taoiseach Micheál Martin at University College Cork during the College of Commissioners’ visit at the start of Ireland’s Council presidency in July 2026.
Photo: PA via Belgaimage.
I. On who writes the rules for training and safety
Governments, rights holders and developers approach the same negotiations with different legal obligations and commercial interests.
28.1: Ireland puts licensing on the agenda as Washington argues for broader training rights
The Dublin copyright conference on 14 October will examine licensing and enforcement under the Irish Council presidency, with NewsBrands Ireland and IRMA as joint hosts and the Department of Enterprise, Tourism and Employment endorsing the event. Its announced speakers include Henna Virkkunen, Peter Burke, Niamh Smyth and Eleonora Rosati, bringing European policy, Irish government and copyright expertise into the discussion. The conference can help define the sector’s priorities; its participants will still need legislators, regulators, courts and negotiating parties to act on them.
In Washington’s account of the problem, AI developers need broader access to training material: Reuters reports that Commerce Secretary Howard Lutnick urged G20 members to embrace the American copyright exception, while the Justice Department supported OpenAI’s position in the New York Times litigation. Those interventions express the US administration’s position and leave the courts to decide the disputes. The G20’s final ministerial statement recognises copyright protection, prior and express consent, and applicable exceptions, while leaving AI-related copyright questions to each member’s established legal processes.
Europe has its own legal basis: Article 53 of the AI Act requires general-purpose model providers to maintain a copyright compliance policy, including respect for rights reservations under the CDSM directive. The sector also has a practical point of reference in SPUR’s work on publisher control, licensing access and standards for usage reporting. Courts establish the legal boundaries, and negotiations establish workable prices; common technical standards should help both parties demonstrate compliance without dictating each publisher’s commercial terms.
Reflections Which licensing requirements should European news organisations bring to Dublin as a common proposal? How can shared technical standards strengthen smaller rights holders while preserving independent pricing and competition?
28.2: Amodei calls for stronger safeguards while Europe already requires systemic-risk controls
The Next Web reports that Anthropic’s Dario Amodei wants developers to allow more time for alignment, safeguards and external evaluation as model capabilities increase. His proposals include outside evaluators with extensive access to company systems, an approach that raises practical questions about independence, confidentiality and authority. In Britain, meanwhile, the government has rejected a proposed national AI “kill switch”, arguing that domestic restrictions would offer limited protection against development and misuse elsewhere.
For the EU, several relevant duties already have legal force: Article 55 requires providers of general-purpose AI models with systemic risk to evaluate models, document adversarial testing, assess and mitigate systemic risks, report serious incidents and protect their models and infrastructure against cyber threats. The general-purpose model rules have applied since 2 August 2025, although Article 111 gives providers of models placed on the market before that date until 2 August 2027 to comply. The AI Omnibus that entered into force on 27 July 2026 postponed high-risk system deadlines, while the Commission’s current implementation guidance retains the general-purpose model timetable. Amodei’s proposed evaluator arrangements require their own assessment; Article 55 establishes substantive duties without prescribing every feature of his proposal.
The practical test concerns the evidence that providers submit: Euractiv reports that the Commission demanded precise information about corrective measures after OpenAI notified it of the DseWiki incident. News organisations procuring agentic systems should ask for documented testing, incident notification commitments and explicit authority to suspend a workflow. A provider’s assurance has operational value when the customer can identify what the provider tested, what failed and who must respond.
Reflections What evidence should a newsroom require before granting an AI agent access to editorial systems or confidential material? How should regulators and customers evaluate a provider’s corrective measures while protecting legitimate security information?
II. On documenting origin and preserving editorial context
News organisations need standards that describe both the material they receive and the decisions they make with it.
28.3: IPTC advances provenance practice as Press Gazette introduces human-authorship checks
The IPTC Photo Metadata Conference examined practical implementations of C2PA, identity assertions and metadata across news and imaging workflows. IPTC also introduced Metawatch, a public service monitoring metadata use across the news industry, including content credentials and AI opt-outs. That work gives the sector a way to examine whether publishing tools preserve the information that standards promise.
Press Gazette has taken a related commercial initiative with Made by Humans, which combines Pangram and Identifai checks in a platform developed with Flip-Pay. Press Gazette says the service checks text, images, video and audio before publication and records reviews for audit purposes. Editors considering such a service should examine its error rates, appeal procedures and definition of human authorship, particularly when a journalist uses AI for translation or editing.
The C2PA specification distinguishes verifiable provenance from a judgement about the truth of the content. A newsroom should make an equally precise distinction between evidence of origin, an assessment of AI involvement and editorial verification. Each answers a different question, and a trustworthy label should tell the reader which question the publisher actually checked.
Reflections What should a publisher promise when it displays a human-authorship or provenance label? How can newsrooms preserve useful evidence of origin while protecting confidential sources and allowing legitimate editing?
28.4: AP and Reuters pursue shared context as suppliers automate more newsroom tasks
Journalism.co.uk reports that AP and Reuters want a common language through which AI tools can exchange context from newsgathering to publication. Buyers should follow the initiative’s development and ask for a published specification and evidence of adoption. The essential question concerns which editorial information every participating tool must preserve.
In its IBC announcement, nxtedition describes a demonstration that uses Reuters material to develop a story, select video, prepare graphics and assemble a programme, with journalists and directors reviewing the output. Separately, Reuters has replaced Getty as ITN’s archive agency partner, with plans to digitise another 10,000 hours of footage and improve discovery through Reuters Imagen metadata and Reuters Connect. These announcements show suppliers combining content, description, licensing and production tools into broader services.
Press agencies have a particular responsibility here as information infrastructure serving many other news organisations. A useful shared specification should preserve attribution, rights restrictions, corrections and editorial status whenever software transforms an item. Publishers and public broadcasters should help define those requirements before procurement decisions make a supplier’s internal conventions their own operating standard.
Reflections Which editorial and rights information must every tool preserve when it transforms agency material? How can customers obtain interoperable workflows without depending on one supplier for content, metadata and production?
III. On discovery, measurement and negotiated access
Audience strategy and commercial measurement increasingly depend on how intermediaries select and represent original reporting.
28.5: Publishers reorganise audience teams around AI discovery
Digiday reports that USA Today plans an audience and digital production team of 23 to 30 people as it responds to changes in search and distribution. At the New York Times, Christine Liang has taken responsibility for AI and off-platform discovery, following the creation of a comparable leadership role at the Washington Post. The organisational choices matter because they assign people and budgets to a changing relationship with audiences.
A publisher needs to know which forms of visibility produce a reader relationship and which mainly improve an intermediary’s answer. Audience teams should therefore compare AI citations and referrals with subscriptions, repeat visits, newsletter engagement and other evidence of sustained value. Optimising for an appearance in an answer carries a different commercial logic from persuading a reader to return.
Public policy also influences discovery: Nieman Lab describes Australia’s proposed Digital Duty of Care, including greater user control over recommendations and personalisation. For European news organisations, that debate suggests a useful test for distribution policy: how much choice can the audience exercise, and what access can a publisher retain? A durable audience strategy should accommodate changing interfaces while continuing to give people reasons to seek out the newsroom.
Reflections Which measures would demonstrate that AI discovery creates lasting value for a publisher? How should audience teams divide investment between intermediary visibility and direct reader relationships?
28.6: Media-intelligence suppliers compete on AI visibility, local context and contract terms
Onclusive’s Unified Platform announcement combines monitoring, measurement and communications workflows, including GEO Analytics for visibility across AI search platforms. The company also describes a shared content foundation that includes directly licensed and paywalled material. Its offer gives communications teams another object to measure: how AI services represent an organisation when users ask questions.
Truescope plans to enter Australia on 1 October with a maximum twelve-month contract term, no automatic renewal and cited AI outputs. In Nigeria, Punch reports that FlashNarrative claims capabilities for local language, code-switching and cultural context alongside its brand-intelligence metrics. These announcements describe the suppliers’ offers; customers still need evidence of coverage, accuracy and commercial performance.
The three approaches suggest a broader procurement agenda: the object of measurement, the quality of interpretation and the customer’s freedom to change supplier. Buyers of AI-visibility analysis should require a reproducible method, including the models, prompts, languages, dates and sampling assumptions behind each result. A metric earns value when the customer can understand its limits and use it to support a defensible decision.
Reflections What minimum methodology should buyers demand before treating AI visibility as a management metric? How should media-intelligence contracts allocate responsibility for source rights, analytical errors and changing model behaviour?
28.7: Indian Express negotiates attribution and editorial control with OpenAI
The Indian Express Group and OpenAI have announced a partnership covering current and archived journalism across seven languages. The publisher says selected summaries and excerpts in ChatGPT will identify the original source and include direct links, while the group retains independent editorial control. The agreement offers a concrete example of a publisher specifying how an AI service should represent its journalism.
The public announcement describes attribution, editorial safeguards and possible uses in archive discovery, research and translation, while leaving payment and detailed usage terms undisclosed. We can assess the announced design and reserve judgement on its economics. For any publisher, the commercial test should include the cost of producing the journalism, the scope of permission and the evidence needed to reconcile actual use with the contract.
European news organisations can use such announcements to identify questions for their own negotiations. Training, retrieval, display, translation and internal newsroom assistance deserve explicit treatment because each permits a different activity. A licensing market will work better when the parties can describe those activities consistently and negotiate the price of each permission on informed terms.
Reflections Which permissions should a publisher price separately when negotiating with an AI provider? What reporting and audit rights would allow both parties to evaluate the partnership beyond attribution and referral counts?
IV. On platforms financing newsroom capacity
Training and technical support can create valuable capabilities, alongside dependencies that managers should assess at the outset.
28.8: OpenAI and Google support journalism training and product development
OpenAI has announced more than 400 ChatGPT Edu subscriptions for students and faculty through CUNY’s Newmark J-School and Medill’s Knight Lab. Its broader programme also includes access and API credits through the American Journalism Project, support for AI engineering fellows through Lenfest, and its WAN-IFRA partnership. The initiatives combine access to tools with the people and training needed to use them.
In Ukraine, WAN-IFRA, AIRPPU and OpenAI plan a Newsroom AI Catalyst programme for ten organisations from 17 September, following masterclasses that began in August. In Latin America, O Globo and nine other newsrooms are participating in an eight-month News Product Alliance programme for Grupo de Diarios América, with Google News Initiative support, to develop AI products for younger audiences. Both programmes address practical capacity, although their announcements offer limited grounds for judging lasting results.
Newsrooms should welcome useful support while defining the capabilities they intend to retain when the programme ends. Managers should establish who owns the resulting tools, which data a partner can access, how staff can transfer their skills and what the service will cost after credits expire. A successful partnership should leave the newsroom better able to make its next technology decision on its own terms.
Reflections Which capabilities should a newsroom retain independently after platform support ends? How should journalism schools and industry associations evaluate the influence of technology sponsors on their curricula and product choices?
V. On synthetic influence and the audience’s ability to judge
Effective safeguards must address commercial incentives, enforcement and the public’s capacity to assess information together.
28.9: Investigations expose synthetic influence while young audiences turn to AI for news
ABC reports Reset Tech’s findings of a network impersonating Australian politicians, with links to an influencer business in Sri Lanka. The researchers found monetisation on roughly one in ten of the network’s Facebook pages, AI labels on only one in three examined posts and deletion in one in 33 cases. Meta told ABC that it was reviewing the identified pages and would act against policy violations. The immediate governance questions concern the incentives for producing such material and the speed and consistency of enforcement.
Semafor reports that YouTube terminated twenty channels after an investigation into “ghost creators”, human presenters reading AI-generated scripts while appearing to offer independent commentary. In Mali, RFI reports the Timbuktu Institute’s findings on synthetic characters and propaganda, alongside its warning about the potential for multilingual recruitment. THISDAY’s examination of Nigeria’s approaching elections adds an important distinction: manipulated political material includes both AI-generated content and older editing techniques. Verification should examine the claim and its context as carefully as the production method.
The audience research gives these cases particular relevance: Western Sydney University reports that more than a third of surveyed Australians aged ten to seventeen use generative AI for news, while fewer than half feel confident distinguishing fact from fiction online. The same study finds that three-quarters believe news organisations have little understanding of their lives. For publishers and public broadcasters, verification and relevance therefore belong in the same strategy. Young people need dependable information that recognises their concerns and gives them reasons to examine its evidence.
Reflections How should platforms demonstrate that their monetisation and enforcement systems reduce the incentive to produce deceptive political content? What would news organisations change if they treated young people’s sense of exclusion as seriously as their exposure to misinformation?
Conclusion
The Irish presidency gives the sector a practical occasion to connect copyright, AI governance, competitiveness and public trust. A useful contribution from news organisations would specify what they need to produce valuable journalism and make it available responsibly. That contribution should include a commercial model, technical requirements and evidence of how the arrangements serve audiences.
A standard resembles the schedule attached to a contract. It defines the fields, permissions and records that determine whether the parties can perform what they promised. If news organisations leave that schedule entirely to others, they accept another party’s assumptions about what matters and what deserves a price.
The Dublin conference on 14 October explicitly links a licensing market with enforceable copyright protection. Publishers, public broadcasters and press agencies should bring proposals that reflect their different roles, including agencies’ responsibility for the verified material that many organisations reuse. Shared definitions of permission, provenance, corrections and usage can support that work while preserving independent editorial judgement and commercial negotiation.
The public ultimately needs to know who produced the information, why it merits attention and who accepts responsibility for correcting it. Europe’s news sector has experience worth contributing to each of those questions. The task for this presidency is to make that experience count in the rules and standards that institutions and companies actually apply.