Publishing teams do not need another vague argument about whether artificial intelligence is good or bad. They need a cleaner way to separate routine workflow assistance from rights uses that still carry real legal and commercial risk.
That is what makes the International Federation of Reproduction Rights Organisations, or IFRRO, worth reading here. In materials it published in April 2026 and June 2026, the federation keeps returning to the same operational divide. It says AI can support some publishing work, but it also says training and large-scale reuse of protected content still require serious attention to authorisation, transparency, enforcement, and remuneration.
The useful takeaway for Rex readers is not that IFRRO has settled the market. It has not. The useful takeaway is that one organised rightsholder constituency is now drawing the line in practical terms: assistive workflow use is one category, unlicensed content ingestion is another.
What IFRRO says counts as a workable AI use inside publishing workflows
IFRRO's consolidated survey report says respondents identified efficiency gains and productivity improvements from AI-assisted processes. It specifically says members referred to AI tools supporting research, summarisation, factchecking, translation only as a helping tool, and administrative processes.
That is a much narrower and more useful framing than the usual "AI will transform publishing" rhetoric. It points to bounded uses that can sit inside a human-controlled workflow: preparing background material, speeding up internal review tasks, helping with discovery, or reducing repetitive admin.
Even there, the report is not saying every use is frictionless. It is saying those are the areas where some respondents can see value if the governance and rights position are handled properly. That matters because many authors and small presses are already facing practical procurement questions: should this tool help with internal research, rights administration, or accessibility support? IFRRO's survey suggests that limited, supervised assistance is where rightsholder organisations currently see the clearest room to move.
Where the hard line starts: training, transparency, and remuneration
The same source set is much firmer on higher-risk uses. IFRRO's public AI page says AI training relies on protected content and, absent an applicable exception or defence, requires authorisation. Its April 2026 governance guide also treats transparency as a precondition for accountability and licensing as a practical route to lawful use.
The survey report explains why members are pressing that point. IFRRO says respondents highlighted concern about unauthorised training use, weak transparency obligations for AI developers, limited opt-out effectiveness, loss of revenue, reduced traffic to content, and growing power imbalance between rightsholders and major AI companies.
That combination is what makes this more than a philosophical copyright debate. For authors, translators, and publishers, the risk is not only that works may be used without permission. It is also that the surrounding market can shift while the rightsholder has too little visibility into what was used, too little leverage over terms, and too little share in the resulting value.
Why the survey matters even if it is not a full market census
This is not a neutral population survey of the whole publishing industry. It is a rightsholder-federation survey, and the article should be read that way. But that does not make it trivial. IFRRO says the survey gathered responses from 24 organisations, including collective management organisations and author and publisher organisations, after discussions with the European Commission on copyright and AI policy developments.
For Rex readers, that gives the report a specific kind of value. It is a good readout of where organised rightsholders think the commercial pressure points are right now.
- Licensing pressure: respondents want workable systems rather than vague promises of future fairness.
- Transparency pressure: rightsholders still see weak visibility into training-data use and output relationships.
- Workflow pressure: members can imagine bounded assistive uses more easily than unrestricted generative exploitation.
- Human-authorship pressure: IFRRO says respondents broadly agree that copyright protection should remain grounded in human authorship.
That last point matters because it keeps the survey from collapsing into a simple anti-technology statement. The federation is not arguing that AI has no place. It is arguing that the system still has to recognise human creative labour and keep fully AI-generated output from quietly swallowing the economic assumptions that authors and publishers depend on.
The mood is cautious for a reason
IFRRO says approximately 65 to 70 percent of respondents consider AI to present more challenges than opportunities at the moment. That figure is not interesting because it proves a universal consensus. It is interesting because of what sits behind it: concerns about unauthorised use, weak remuneration, insufficient transparency, and the lack of an effective enforceable framework.
The report also says a minority of respondents expressed strong scepticism about generative AI in fields such as literary translation, citing concerns about quality, loss of creative control, and degradation of professional standards. That is a useful nuance. The same report that allows for AI-assisted research or administration still shows real resistance when AI starts pressing more directly on authorship, translation quality, and market substitution.
The practical takeaway for Rex readers
If you are an author, translator, rights manager, or small publisher, the cleanest lesson from IFRRO's 2026 materials is to stop talking about AI as one single use case.
Ask narrower questions instead. Is the tool helping with internal research, factchecking, accessibility support, or administration under human supervision? Or is it consuming protected content in ways that trigger licensing, transparency, and remuneration questions you cannot answer cleanly?
That distinction will not solve every contract or policy problem. But it is a better operating model than either blind enthusiasm or blanket panic. It also fits with our translation rights checklist for authors and our guide to copyright recordation after rights transfers: rights discipline gets more valuable, not less, when new tools start rearranging who uses content and how.
If you need help tightening rights, translation, or publishing workflow before AI policy confusion turns into a licensing problem, contact Rex Publishing.