What publishers do when everyone has AI

Dustin and BMJ’s Ian Mulvany discuss editorial judgment, research curiosity, agent-ready publishing, and where they might be wrong about AI.

blog.hum.works

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

And that's a wrap from me (I'm about to run out of battery, but I've made it throgh the day) and what a fantastic day it's been - a great great day, I've really enjoyed it. I'm closing out the live blog now.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

The closing remarks are optimistic, expectations for next year content indispensable to AI innovation in peer review research integrity a glimpse into the next chapter of this industry more new ideas new disrupters emerging some topics we will still be talking about

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Close to closing remarks I really like Caroline's point that we are uniquely placed to help the creation of AI that is good for science. Stephen - huge potential commercial upside. Jess - new commercial models, but ones that we can communicate, and models that are flexible. Laura - the way content…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Caroline has a call to try to avoid fragmentation. She wants to identify the areas of collaboration. Laura - trust, integrity, TRACE, things we can't solve as individuals. Atul - simple standards. Barbara - making sure the content is discoverable - we make impact possible. Jess - the systems are aw…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

So it's going to come down to culture. AWS were great at creating small pockets of innovation - even in a behemoth of an org.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Question from Stephen - will ai accentuate the scaling inequalities we have in the industry, or will it break them making it easier for small publishers to survive better? (IMO on the one hand AI gives the power and agility to small teams, but larger organisations can get to market at scale). Atul…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Do they think we will need new roles? Or same roles augmented. Basically - yes to both. That's the story on that question. That's it.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

We need to give explicit permission to get people to take risks.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Question - is this a war time moment for these companies, or are they going to remain more collaborative - For Laura people are struggling with the idea that they can take more risk. Atul describers it as a fundamental workforce transformation process. Atul does not think we are in war time - it's…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Barbara mentions the growing voices around AI worries - (a good link on this is - this one - https://www.writingruxandrabio.com/p/curing-cancer-wont-redeem-ai). She asked an ethicist to come in and help them situate their concerns. He talked to them about purity and ethics - a really interesting pe…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Caroline asking about AI and what might be areas of disruption. Jess talks about the burden on peer review, and the increasing volume of requests coming in. So this is a core place. There is an opportunity to rethink roles, how might we make editor and publisher roles better. Laura is seeing the sa…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Caroline asks - who does not trust us - for me I think one thing that we might need to think about is that agents are not cognizient and so they need explicit signals, rather than implicit signals (I could be wrong about this because the LLM has enough world knowledge, but that world knowledge curr…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Ok, so we have an agreement that just is important, but given that no one is going to say that we need to be moving towards a low trust ecosystem, and a lower trust relationship with our communities, which means that there is a sense in which this question is potentially an empty question. It might…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

On the question of operationalising trust: Stephen - for him it's about the process, and creating a high quality process. Barbara cites Susan Fisk - trust coming from a combination of warmth and competence, and the warmth is the related to the intent. For Jess it's an imperative as it is connected…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Well - last session - here we are!! Featuring: Atul Goel, President, KnowledgeWorks Global Ltd Laura Hassink, Managing Director/President for STM Journals, Elsevier/RELX Steven Heffner, Chief Publication Officer and Managing Director, IEEE Barbara Kline Pope, Executive Director, Johns Hopkins Unive…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

I ask the panel about whether they are optimistic or pessimistic about the future - optimistic, optimistic, optimistic, realistic, realistic. Rob thinks that there is going to be some pain.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

This was long but good. Now we have 15 mins for questions. I can't think of anything to ask.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Kim - thinks the big decision we need to make are submission fees - need to increase the friction to manage the incredible growth. Witold - thinks the most important is to train libraries so that they can train their community - PhDs who then train their professors, basically increase knowledge. Bu…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Taking now about what guidelines should be created Frontiers have created an AI playbook - https://www.frontiersin.org/ai-playbook/home - this is a living document - right now for peer review it says "Humans remain accountable: AI can support process but must not be the decision-maker or the source…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

We now enter a bit of a cross discussion on the panel - the question on retraction, and being able to tell the story of the paper - comes up. There is agreement that this does not happen today but that there is value in giving these agents context.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Some discussions on what we might need to help with all this mess - and my message on this is that things are going to get a lot weirder before we can consolidate our ideas about the right future model.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

The arXiv policy - https://blog.arxiv.org/2026/10/01/updated-rate-limit-policy/ - has come up. Kim points out that while the principle is interesting, it will likely be hard to enforce. Kim mentions submission fees. She would love quotas but she would love institutions to be more involved in these…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Jonathan - why not start with the idea that AI can do everything, and work backwards from there. Rather than the other way around. This is what you might do if you started as an AI native organisation. For internal facing things they have a high tolerance to blast radius as they have a small organi…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Oh, an interesting point - first time it's come up - and something we have talked a lot about in BMJ - AI can not only help against slop - we could - and we should - use these tools to make papers better - there are far more papers that AI could be used to help - and the upside is so much larger wh…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Kim points out that AI can help us with automated rejects, e.g. if there is a paper with a duplicate submission. That is not asking it to automate judgement, but automating where we have hard guidelines already. (EXACTLY!).

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Jonathan - the systems their company is building is aiming to help publishers understand how these systems work.

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Next Jonathan - who will be talking about agentic consumption infrastructure. all of their focus is on transparency in this context. Jonathan points out that the LLMs are narrating out what they are going to do (I heard that interacting with non RFLD's non aligned LLM is like speaking to a mad wizz…

blog.mulvany.net

Live blog: STM - Frankfurt 2026 - Knowledge Infrastructure Responsible AI

Maria - humans need to be in control of final decisions. (I disagree with this, I don't think it needs to be applied in every case). So they are moving away from tick boxes - they want the handling editors to provide comments, not just box ticking. They are making some policy changes around things…

blog.mulvany.net