Certain interactive features from this article are only available in the full digital issue.

The conference is also covered on independent blogs by Elisabeth Bik and Anna Abalkina.

Read Full Issue (subscribers)

In This Feature

• Opening Ceremony

• Misuse of AI by Paper Mills

• GenAI and Research Integrity: Global Ethics, African Wisdom, Shared Responsibility

• Championing Research Integrity: Insights from the Einstein Foundation Award Winners

• Research Security and Integrity

• The Steneck-Mayer Lecture: Science under pressure in a fractured world

• Next Stop, Tallinn

• Voices From the Program

Opening Ceremony

Three Forces, One Forum

Held every two years on a different continent, the World Conference on Research Integrity (WCRI) has become the central forum where the research community gathers to confront the questions shaping how science is done and trusted. In his opening remarks, David Moher, a clinical epidemiologist at the University of Ottawa, reported that the sold-out program had been drawn from 814 abstract submissions and 51 workshop, symposium and focus track proposals, reviewed by 68 peer reviewers.

Moher shaped the program with fellow co-chairs Chris Graf, Director of Research Integrity at Springer Nature, and Lex Bouter, founding chair of the World Conferences on Research Integrity Foundation (WCRIF), who is also a member of the Science Integrity Alliance (SIA) Advisory Board. The three conference themes addressed the forces and questions reshaping research integrity today:

1. Artificial Intelligence (AI)

Framed AI as both a powerful tool that can improve research quality and a force that can obfuscate research integrity and enhance questionable research practices.

2. Research Security (RS)

Examined the inherent tension between protecting sensitive information and maintaining the traction of open science.

3. Indigenous Knowledge Systems (IKS)

Invited the integrity community to engage more deeply with Indigenous ways of knowing, being and doing, and to reflect on how these perspectives can expand and reshape understanding of research integrity.

In the welcome addresses, Mona Nemer, the Chief Science Advisor of Canada, proposed a Science Compact, guiding principles agreed not among nations or institutions but among individuals, since integrity cannot be delegated or outsourced. She also shared that all Canadian federal departments that conduct or use science have adopted a model science integrity policy covering this year's three WCRI themes. Alejandro Adem, President of the Natural Sciences and Engineering Research Council of Canada (NSERC), spoke on behalf of Canada's three federal research funding agencies and framed openness and security as essential pillars of research integrity rather than competing priorities. He also named IKS as integral to the work ahead, noting that scientists and engineers have much to gain from the thoughtful and respectful decolonization of research.

Mona Nemer, Chief Science Advisor of Canada

The Difficult Truths

The final welcome address came from Ry Moran, a member of the Red River Métis, Inaugural Associate University Librarian for Reconciliation at the University of Victoria and founding director of the National Centre for Truth and Reconciliation. He spoke on difficult truths, reframing research integrity as a deliberate act of repair, restoration and resurgence, not a new form of old demons. Drawing on the late Justice Murray Sinclair's definition of reconciliation as building and keeping mutually respectful relationships, Moran put two questions to the audience: how the human family will choose to live together in respectful ways, and how to live with the planet itself. Quoting Anishinaabe Elder Dr. Harry Bone, Moran reminded the conference that Indigenous peoples have been researched to death, and the time had come to be researched back to life. He closed with a question that echoed across the four days: If our research does not help us live more respectfully with one another and with the planet, what and whom is it actually for?

Misuse of AI by Paper Mills

The opening plenary addressed the misuse of AI by paper mills, companies that generate and sell fake manuscripts to researchers, preying on academia's perverse incentive structures by churning out fraudulent science for profit. Session chair Jennifer Byrne (Conjoint Professor at the University of Sydney and a member of SIA's Advisory Board) welcomed Daniel Moreira of Loyola University Chicago, Stefan Stender of Copenhagen University Hospital and Renee Hoch of the Public Library of Science (PLOS).

Moreira traced the field from the lone wolves of the past, where a single researcher might tamper with Western blots across a career and get millions of dollars of funding, to today's "perfect storm" (see box). But not all is lost. Moreira pointed to emerging defenses: image provenance analysis to track image reuse across papers, synthetic image detection that reads background noise to tell whether an image is fabricated, as well as community competitions on copy-move detection. One such competition, recently run by Moreira with SIA partner recod.ai, drew more than 1,500 teams and awarded $55,000 in prizes.

Paper mills apply generative AI (GenAI) across the entire research pipeline:

• synthetic Western blots produced in two months using off-the-shelf machine learning solutions

• AI-generated peer reviews with no human oversight

• fabricated manuscripts built only to survive a quick glance

Western blots are a common lab technique for detecting proteins. Because the results appear as images in papers, they are a target for manipulation.

The 20-Minute Manuscript

Moving away from images, Stender used Mendelian randomization (MR) as "a canary in the coal mine story" for a development now seen across other open-access datasets, a flood of mass-produced papers exploiting publicly available data. The method uses inherent genetic differences to test whether a modifiable factor helps cause a health outcome, decreasing confounding and reverse causation problems that traditionally affect observational studies. In Stender's words, easy access to GWAS (genome-wide association studies), user-friendly statistical packages, and publicly available data have democratized the use of MR, which facilitated the mass production of papers, from 100 to 200 annual PubMed hits a decade ago, to more than 7,000 in 2025. This increase was solely driven by papers from Chinese researchers, almost all of them using MR on publicly available data. This is easily done, and was shown in his demonstration that Google Gemini produced a full paper with the appropriate sensitivity analysis from a one-line prompt to write an MR paper on noodle intake and educational attainment.

Stefan Stender, Copenhagen University Hospital

The Detection Arms Race

Hoch presented PLOS data showing a sharp rise in large-scale paper mill cases that began with articles published in 2020, spiked again in 2021, and have been maintained at a pretty high level since (see graph below). She argued that the total impact is underreported, since their analysis only includes post-publication issues for which formal investigations are completed or nearly complete. PLOS has responded to this threat by increasing its publication ethics team, adding to PLOS' preventive pre-publication integrity screening program, and introducing additional pre-review editorial checks. These efforts aimed to improve PLOS' early detection of problematic submissions, and resulted in a major rise in PLOS One's desk reject rate since 2021. PLOS also piloted AI text detectors. Early results suggest that a simple yes-or-no on AI-generated text does not distinguish paper mill submissions from random ones; however, score distributions show some differentiation, indicating that AI text scores are "a helpful risk indicator, not a perfect litmus test, but a risk flag." Hoch noted that AI makes paper-mill content easier to produce and harder to detect. The escalating arms race she presented was captured visually by Pedro Veliça in his comic for the January–March 2026 issue of REACH.

GenAI and Research Integrity

The second plenary, “GenAI and Research Integrity: Global Ethics, African Wisdom, Shared Responsibility,” brought together three scholars to closely examine how the field's dominant frames for AI ethics look when set against historical record-keeping, Yoruba virtue ethics, and African research integrity governance. Chaired by Moher, the session featured Lorna Waddington, Associate Professor of International History and Academic Lead for Academic Integrity at the University of Leeds, Helen Titilola Olojede, Lecturer and Head of the Department of Philosophy at the National Open University of Nigeria, and Retha Visagie, Research Integrity Manager at the University of South Africa, co-founder of SARIMA's (Southern African Research & Innovation Management Association) Northern Regions Community of Practice, and Chair of the EthiXPERT Board, a SIA partner organization.

The Presence of Absence

A self-described "very pro-GenAI" scholar, Waddington said she was consistently censored across major large language models (LLMs). Two years of empirical work, including a decolonized genocide studies curriculum she co-developed with a First Nations scholar, had taken her into what she called "the presence of absence," a scenario where knowledge is produced by GenAI but the user does not know what has been missed out. Sometimes, key areas are entirely ignored. She documented multiple categories of failure, including hallucinated authorship of Nazi-era propaganda followed by links to extremist sources, refusals framed as capacity claims ("I don't have the capacity to do this") rather than policy choices, and a fabricated rationale for refusing to surface a German soldier image in 1940, while complying with the request when the year was 1941. She also argued that AI training data is a curated selection shaped by decisions about what's worth recording, what's technically possible to record, what's in languages with large digital footprints, and what's deemed credible.

Lorna Waddington, University of Leeds

Asked about effective forest management, the leading LLMs showed a three-move sequence: first, they omitted Indigenous practices; then, when challenged, they agreed it was Western bias; and when pressed again, they still left it out. Waddington called this "governed suppression" and noted DeepSeek replicated the same bias.

The Omoluabi Researcher

Olojede anchored her talk in Omoluabi, a West African virtue ethic from the Yoruba people that defines a person by the integrity of their character in relation to others. She explained that research ethics has been largely based on compliance, focused mainly on rules, sanctions, similarity checks, and detection machines because these are easier to track and measure. Instead, she proposed supplementing this with researcher integrity drawn from Omoluabi's core virtues: Iwa lewa (character), Otito (truth-telling, honesty), Irele (intellectual humility), Akinkoju (courage), respect, and Iwontunwonsi (balance, moderation, “not too much and not deficient”).

An Omoluabi researcher, she said, does not ask whether something will pass peer review, but what the research says about them as a person in their community. Answering the common objection that people do not hold steady character traits over time, she said becoming virtuous takes sustained work, which is why, as she put it, “virtue is a character trait and not a habit trait.” She added that this must run from early childhood through university and be embedded in supervision, evaluation, and AI systems as integrity-by-design, rather than damage control or integrity washing.

Titilola Olojede, National Open University of Nigeria

Toward an African Code

As Visagie took the podium, she positioned herself as a non-Indigenous scholar speaking from a situated African perspective, but not for all African stakeholders, and as an ally. Her talk was a call for an African code for research integrity, inclusive of responsible AI practices and grounded in epistemic and temporal justice. Such a code, she argued, is no longer optional, because AI's impacts are broad, deep, and hard to foresee, and rapid advancements have exacerbated structural strains in African higher education, requiring more agile governance framed through a developmental lens. Visagie cited a qualitative study where Global North students saw GenAI as complementary, while Global South students faced what she called invisible labor, working harder for AI to pick up their tone.

She proposed a move from compliance to a temporal epistemic justice model in which accountability shifts from the person-centered Western framework to a relational epistemology where knowledge is co-produced with the communities and elders; within it, the impact is measured not at the moment of publication but by the long-horizon consequences of knowledge in 30 or 50 years. Thus, she argued that Indigenous and community-based methodologies must be recognized as rigorous.

As African codes focus on reciprocity and solidarity, the question is no longer whether researchers are complying but, in her words, “whose knowledge will actually count and whose futures should be impacted?” While she had initially thought Ubuntu, a southern African ethic centered on shared humanity, would be her answer, quickly came the recognition that Africa is not one thing. Ma'at (an ancient Egyptian tradition), the Sankofa bird (a Ghanaian symbol), and West, East, and Central African ethics all sit alongside Ubuntu, and Ubuntu's appearance in UNESCO documents has not closed the implementation gap. Her proposal was for convergence rather than replacement, integrating the principles of reliability, honesty, respect, and accountability in the European Code of Conduct for Research Integrity with African ethical traditions and epistemic justice to help develop a continentally acceptable AI research integrity framework. She closed with a real-world case: a non-native English-speaking postgraduate student, repeatedly flagged by the AI detector despite having openly declared her use, who faced a one-semester delay in graduation while she was under investigation. Her point, she said, was not against AI detectors but for greater and more careful reflection on their algorithmic biases.

Championing Research Integrity

For the third plenary, "Championing Research Integrity: Insights from the Einstein Foundation Award Winners," three of the award's winners took the stage to work through reproducibility, post-publication review and the conditions that let misconduct persist. Mai Har Sham of the Chinese University of Hong Kong, a jury member for the €350,000 award, chaired the plenary session and introduced Tim Errington, Senior Director of Research at the Center for Open Science (COS), a SIA partner organization that won the 2021 Institutional Award; Brandon Stell, neuroscientist at France's CNRS and co-founder of PubPeer, which won the 2024 Institutional Award; and Elisabeth Bik, microbiologist and independent science integrity consultant, who won the 2024 Individual Award.

Five years in the spotlight

REACH profiled the Einstein Foundation Award and its 2021 to 2025 winners in an earlier Spotlight feature by Ulrich Dirnagl and Ulrike Pannasch.

Read the Article

Errington opened his talk by acknowledging that research exists to produce generalizable knowledge and repeatability matters only because it helps us contribute to that knowledge (see box). Funded by the US Department of Defense, the COS' seven-year Systematizing Confidence in Open Research and Evidence (SCORE) project looked at those questions across the social and behavioral sciences, from criminology to sociology, sampling 600 papers from 62 journals. The point Errington was most excited about was not the findings but that more than 800 researchers had done the work. Across the sample, results were often difficult to reproduce, far more so when authors had not shared their data and code. Once the team moved from the same data to new data, statistical significance became close to a coin flip and replicated effect sizes shrank on average by 58%.

The 3 Rs of Repeatability

1. Reproducibility: the same data and the same analysis returning the same result.

2. Robustness: the same data analyzed a different way, returning the same result.

3. Replicability: new data collected to test the same question, returning the same result.

He insisted that researchers continue to do state-of-the-art work, as inefficiency is exactly what to focus on going forward. At the root of it, he argued, is what we reward: whether novel positive results and tidy stories that allow for selective reporting and other questionable research practices that no one is held accountable for, or true innovation. The answer is to shift the whole culture rather than just ask researchers to show more of their work. He laid it out as a reinforcing pyramid, in which a policy at the top fails without the infrastructure, ability, incentives and norms beneath it, just as infrastructure built without rewards goes unused. Errington's reproducibility findings were only a slice of what SCORE produced. The full SCORE figures, which Errington and colleagues set alongside related results from the META-REP program and the Institute for Replication, appear in this issue's RE:CHECK.

Tim Errington, Center for Open Science

Stell traced the motivation to start PubPeer to his time as a graduate student at UCLA, where journal clubs left every paper with some problem, often a minor one, yet offered no forum to take that scrutiny public, so he set out to build one. To show why that mattered, he pointed to the Wakefield case, the 1998 paper claiming a link between vaccines and autism that the research community discredited quickly, but which took a decade to retract. In the years between, parents cited it as a reason not to vaccinate their children, and UK figures over that time period show measles vaccine uptake falling as measles cases climbed.

Complementing Errington's presentation on poor reproducibility, Stell cited two pharmaceutical companies that had tried to reproduce published findings and largely could not, even though both had a commercial reason to want the results to hold. The deeper trouble, he argued, is an incentive system that rewards quantity over quality. A young researcher watching the data trend toward a career-making paper published in Nature has every reason to keep going, even as the outliers pile up. PubPeer's answer was to build an evaluation layer through expert assessment of the data itself, centralized and searchable, that hiring and promotion committees might weigh alongside citation counts, and which might give a researcher pause before steering a result toward that Nature paper, since it could end up there.

Brandon Stell, PubPeer

An Eye for Duplicates

A microbiologist by training, Bik opened by recounting how her sleuthing work had begun with a search for plagiarism of her own writing, and by noticing a weird skill for spotting duplications in Western blots (see image below). She described scanning 20,000 scientific papers by eye and finding around 800 (4% of the set) with image duplications, roughly half of which she judged as fraudulent. The duplications take a few forms: whole panels repeated exactly, panels that overlap, or elements copied within a single image, sometimes mirrored, rotated, or stretched. She reported those 800 papers in 2015 expecting quick action, but a decade on, papers from that same set are still being retracted. She has now flagged close to 10,000 problematic articles, a substantial body of work that has resulted in 1,600 retractions.

She named paper mills and GenAI as the threats she sees rising, noting that a study by Byrne and colleagues estimated around 10% of the cancer literature carries paper mill characteristics, and that AI now produces Western blots realistic enough that her own eye can no longer be the test. Bik's answer is to turn the problem around, so that rather than proving what is fake, the community has to verify what is real. She perceives misconduct rising, rewarded, and costly, since gaming the metrics still buys a better career. In contrast, the critics who expose fraud pay in harassment, lawsuits, and a heavy personal toll, mostly without institutional protection (for another sleuth's first-hand account, see the Chronicles section). Even so, Bik stressed that most science is still sound. She closed with a list of changes she wants to see:

• rewarding accuracy

• hiring replicators

• crediting negative results

• verifying ethics and data

• enforcing consequences

• protecting whistleblowers

An award winner gives back

Bik won the 2024 Einstein Foundation Individual Award

and gave her prize money to the Center for Scientific Integrity to fund other sleuths to attend conferences and purchase detection software.

Research Security and Integrity

The fourth plenary, "Research Security and Integrity," took up a theme that a shifting geopolitical climate is increasingly shaping: how politics now reaches into research integrity, open science, and security. Chaired by Joeri Tijdink, board member of the WCRIF, the session brought to the stage Carthage Smith, head of the OECD Global Science Forum, Sakine Weikert, head of section at the German Academic Exchange Service (DAAD) and its Center for International Academic Cooperation (KIWi), and Richard Gold, a Distinguished James McGill Professor at McGill University and Director of its Centre for Intellectual Property Policy, to set out what research security is and whether open science can underpin global security.

Security Without Bureaucracy

Smith asked whether the global research ecosystem will thrive or struggle, with integrity and security both critical to its state. He defined research security as preventing undesirable foreign state or non-state interference with research and the misappropriation of research outputs, to protect national and economic interests, and argued that it is very much part of research integrity. When he asks government ministries he works with about their science, technology and innovation priorities, security and competitiveness come first, but much of the security talk is really about economic competitiveness. The threats reported are a mix of traditional espionage, coercion, foreign influence and theft, and the hard part is a big gray area where it is unclear what counts as a security risk. Research integrity is mainly held by the scientific community, he noted, while research security pulls in many ministries, from Defense to Economics, that do not all really care about academic freedom and open science. The challenge is then to mitigate the risks proportionately, while still keeping research free and collaborative. The 2022 principles from the OECD, the G7, and the European Commission warned against turning that challenge into a new bureaucracy. By Smith's own account, that is exactly what has happened.

Carthage Smith, OECD Global Science Forum

Even the most basic question regarding what actually needs protecting is left without a clear answer. Smith noted that technology readiness levels, long used to mark basic research as open and applied research as protected, are very hard to define in fast-moving areas like quantum and AI, where it is about the whole field and who gets ahead. Drawing on OECD's STIP Compass (box), he reported a tenfold rise in research security policies and a fourfold rise in the countries adopting them over seven years, with the move from strategies to regulations, high-risk lists, and tighter conflict-of-interest rules. Smith's main point was that research security should be seen as a critical aspect of research integrity, overlapping with it through due diligence, conflicts of interest, and peer review, so the community owns it rather than having it imposed from outside. Overzealous measures can stifle science and actually make a country less secure, and he warned there is a need to create ethical frameworks to guide the building and use of databases of individual researcher profiles to cross-link their publications, funding, institutions, and even their social media. He closed his talk by calling for the adverse effects of recent security measures to be monitored closely, as some countries are already starting to do.

The Right Balance?

Where Smith spoke of research security, Weikert preferred the term knowledge security. She said international academic cooperation is no longer just exchange and mutual benefit, but has become increasingly bound up in a mix of strategic, political, and economic contexts. By her account, the policy push is fast. She complemented the shifts Smith laid out with a set of recent steps from the European Union (EU), such as the 2024 Council Recommendation on research security; the first European flagship conference in Brussels in 2025; the EU's first Research Security Monitor; research security set to be written into the upcoming European Research Area (ERA) Act; and plans for a European Centre of Expertise and a due-diligence tool.

Weikert traced Germany's own steps, including a National Security Strategy, a China Strategy, and a 2025 position paper urging the creation of a national research security platform. There, academic freedom is explicitly protected by the constitution; to her, knowledge security should support it, with trust, transparency, and reciprocity defined in advance. Such a supportive approach runs through KIWi, the advisory unit Weikert heads at DAAD, set up nearly seven years ago and advising German universities on about 600 cases a year. DAAD itself works in 70 countries through regional offices, so the advice can draw on local knowledge and stay specific to each country, even where the tools are the same everywhere. The cases her unit handles show what that looks like in practice (see box).

Weikert listed tools for handling cases like these, from DAAD's 2020 No Red Lines report to a recent KIWi checklist including 17 guiding questions, developed to start a dialogue and work through gray areas rather than rule cases in or out.

Openness and Security

Gold set out to flip the script, arguing that open science is essential to research security rather than a threat to it. He started from Smith's own definition, research security as a matter of economic prosperity, which a country can protect only with access to the knowledge, technology, and know-how it needs to respond to a crisis. Borrowing Europe's own idea of technology sovereignty, the drive to lower a country's dependence on foreign actors, he held that most countries do not have control over the intellectual property and knowledge they rely on.

Framing resilience as the capacity to respond quickly, Gold credited it with getting the world out of the COVID-19 pandemic through the fast development of mRNA vaccines. He argued middle powers could not achieve that alone, since technology sovereignty, despite its national focus, depends heavily on cooperation. His evidence was patent data, showing that countries like Canada invent plenty but do not own what they make: only 10% of recent patents go to Canadians, 90% to foreigners, and 47% to the US alone, which to him meant dependence, not independence. He returned to his opening claim, that open science is the answer, keeping a country from being locked out of the knowledge it needs, since no one can exercise patents, cancel licenses, or use their own legal system to block access.

It also lets knowledge move fast when a crisis hits, so a country can put together a response and adopt the technology it needs in time. Gold argued that had open science been in place during the COVID pandemic, a hub in South Africa could have produced a vaccine about 18 months sooner, but pharma companies refused to share, so it had to be reconstructed. In contrast, he mentioned the world-leading drone industry built by Ukraine in just 12 years through openly sharing knowledge within its own community, and now exporting it to the Middle East. His closing warning was to be skeptical of anyone who says the answer is to wall off knowledge and data, since that only under-powers the research a country could do and leaves it more dependent.

Richard Gold, McGill University

The Steneck-Mayer Lecture

The Steneck-Mayer Lecture invites an original voice to take on one big question. Named for Nick Steneck and Tony Mayer, who founded the WCRI in Lisbon in 2007, it honors the pair whose 2010 Singapore conference produced the Singapore Statement, which is still cited in many codes of conduct. Co-chair Bouter introduced the lecture framing its aim as an original and inspiring contribution on an issue that matters to the research system. This year's invited lecturer was Robbert Dijkgraaf, a theoretical physicist and mathematician who has led the Royal Netherlands Academy of Arts and Sciences, directed the Institute for Advanced Study in Princeton and served as the Netherlands' Minister of Education, Culture and Science, and is now President-Elect of the International Science Council (ISC).

Robbert Dijkgraaf

Speaking under the title "Science under pressure in a fractured world," Dijkgraaf set aside the What of research integrity for the Why. He borrowed a line attributed to Leonid Brezhnev, who summed up the Soviet Union in one word as "good" and in two as "not good," and offered it as his own verdict on the state of science. Then he started with the good. This century has learned to build with the building blocks the 20th century discovered, he said, through genetic editing, quantum and nanomaterials and AI. He likened earlier science to the old maps where cartographers drew sea monsters across the blank spots; the work now is less about exploring nature than about what we can make. AI should help find structure in the "messy middle" of life, behavior and thought, and a worldwide pool of talent is joining the work as whole parts of the world come online.

Then he turned to the pressures. He said academic freedom is sliding under attack, the targeting of research infrastructure in war and the censorship of sensitive areas, while disinformation now comes "often disguised as science." Governments increasingly read research through the lens of economic competitiveness, and science is moving from the public to the private sphere. In the US, the ratio of public to private funding, once roughly one to two, has shifted to one to four, with the top five tech companies set to spend close to $700 billion this year. Diplomacy is under strain too. The United Nations (UN) and its agencies are being undermined, international cooperation is reframed as a security risk, and the Global South is losing faith that industrialized nations will ever deliver on the promise of science.

His own reading was that the backlash is a consequence of success, with science delivering uncomfortable truths and drawing an equal pushback in return, the social equivalent of Newton's third law. The COVID pandemic was a powerful accelerant. He had expected the vaccines to be science's "Sputnik moment" and a wave of public investment; instead came real attacks on trust. His sharpest worry was polarization along educational lines, confidence in universities and the press splitting from confidence in the police, the military and religion.

Robbert Dijkgraaf

Dijkgraaf said the harder news is that many internal pressures are self-inflicted, which is why they can be fixed. The incentives reward "quantity over quality," a tyranny of metrics that falls hardest on young scientists. He recalled a Princeton colleague who held the highest h-index in the physical sciences and, when told, had no idea what the index was, a case of Campbell's law, that any indicator, once set, begins to corrupt what it measures. Overstatement is part of it; his favorite example was the colleague who switched on the Large Hadron Collider and, asked the odds of creating a world-swallowing black hole, answered "small," then had to face a committee of worried Geneva residents. Fraud, fabrication and the replication crisis have grown over twenty years of these conferences from small early cases into serious storms. Trust comes on foot and leaves on horseback, he reminded the room, citing the tobacco industry's "doubt is our product" as proof of how cheaply manufactured uncertainty undoes real findings. Peer review is near breakdown, he warned, and AI-generated output is rising in an arms race no one has yet answered.

Dijkgraaf then turned to solutions: reforming hiring and incentives, overhauling peer review, sharing data and code, investing in replication and educating early. We drive safely, he noted, not because a policeman stands on every corner but because of traffic rules, licenses and traffic lights, and science needs the same. On AI, he urged proactive, clear policies, and singled out the Vancouver Standard for AI Disclosure in Research, proud that the ISC is among the bodies developing it (box).

He turned last to research security, where science is "securitized and even weaponized." With almost every key technology having a dual use, international cooperation is increasingly read by governments as a risk. The path runs between two cliffs, "one is to be paranoid and the other one to be naive." Dijkgraaf said that governments tend to be paranoid about scientists being naive, while in turn scientists are naive in thinking governments are only paranoid, and the two need each other. None of it can be judged without deep domain expertise, because no one in government or intelligence can assess an intricate part of a quantum computer.

To make the point, he then turned to the Princeton office he kept for ten years, once Robert Oppenheimer's and unchanged since. In the 1940s and 1950s, Oppenheimer, Einstein and von Neumann worked almost next door to one another, all three shaping the development of nuclear arms yet taking very different public roles. Einstein became, with Bertrand Russell, a father of the peace movement. Von Neumann, "the coldest of cold warriors," when asked whether he really wanted to bomb the Soviet Union tomorrow, is said to have answered "no, this afternoon." Instead, Oppenheimer took the middle ground, and in the public imagination he was wrong at both ends, "both a communist and a war criminal." The three, Dijkgraaf suggested, mark the narrow path a scientist has to walk on security.

Scientists should actively push back against the pushback and defend academic freedom rather than retreat into a defensive, risk-averse caution, since international collaboration is in a nation's own interest. He described science as an ecosystem and a "global public Commons" the community creates and tends, whose best qualities are "emergent," belonging to the whole rather than to any one scientist or institution. To study the ocean you can look through a satellite, he said, but you still need the local fishers to tell you what is happening in that stretch of water.

Dijkgraaf concluded his lecture by arguing science is in transition toward a better system whose ingredients, diversity, open science, public engagement, security and diplomacy, are still being worked out. Politicians sit in the first ring of the geopolitical theater, the "splash zone," while scientists sit a ring back, able to take initiatives and outlast electoral cycles where a minister these days averages under 22 months. Open science and free exchange are "not an ideological view, it's an operational requirement" because they give better results. The real question, with more people and tools in science than ever, was whether the community has the imagination to build something better.

Next Stop, Tallinn

On the final afternoon, delegates gathered once more at the Bayshore Grand Ballroom for the Closing Ceremony. Conference co-chair Graf opened by thanking the room for four days of engagement and contributions, before Tijdink took over, speaking on behalf of the WCRIF and serving as master of ceremonies, alongside Byrne, for the conference awards (see the list of awardees).

The ceremony then turned to those who built the conference series. Board member Zoe Hammatt led a remembrance of the late Susan Zimmerman, former governing board member and a formidable champion for research integrity in Canada and beyond, before the three retiring members of the WCRIF were honored. Sabine Kleinert had been part of the series since its earliest planning in 2006, co-chairing the 3rd, 4th and 7th conferences, holding out for scholarly rigor and bringing publishers into a wider conversation. Bouter, who back in 2014 turned up to meet WCRI co-founder Nick Steneck with a suitcase of Gouda cheese and a long list of questions, created the Foundation in 2017 and co-chaired the 5th through 9th conferences. Lyn Horn, thanked in absentia, brought a sense of social justice to the board as a founding member of ARIN and led the Cape Town Statement on fairness and equity. Graf closed out the thanks, to sponsors and exhibitors, co-chairs, the committees, reviewers and judges, the local team, the organizers Global Planning Solutions, and Elder Bob Baker and hoop dancer Jay Genaille from the opening.

Then came the drumroll the room had rehearsed. The 10th WCRI will be held in Tallinn, Estonia, in 2028, the first time the conference travels to Northern or Eastern Europe. The Tallinn conference will be supported by the University of Tartu, the Estonian Research Council, the Estonian Academy of Sciences, the Estonian Convention Bureau and the City of Tallinn, and co-chaired by Hammatt and Graf for the WCRIF. Dates, venue and program are still to be announced. Margit Sutrop, Head of the Centre for Ethics at the University of Tartu, member of the Estonian Parliament and co-chair of the upcoming conference, took the stage and thanked the Vancouver team, promising a warm welcome, sea views and a medieval Old Town that is a UNESCO World Heritage site. Vancouver holds a deep personal meaning for her, she said: her grandfather, a sea captain she never met, lived in the city, and in 1954 his ship sank on the way to Alaska. During the conference she brought flowers to the ocean in his memory. Moving from Vancouver to Tallinn felt to her like moving from one city by the sea to another. After describing Estonia as a small country and a singing nation that sang itself back to freedom in 1991, she closed with an invitation:

Margit Sutrop, University of Tartu

Voices From the Program

The conference ran wider than its main stage.

Workshop leaders, presenters, award winners and conference organizers shared what they brought to Vancouver and what they took from the largest in-person global forum on research integrity. Their reflections, along with the week's award recipients and comments from more voices, are gathered here. Fuller interviews and additional comments are available to SIA subscribers in the HIKE forum.

Jana Christopher treats a microscopy image as evidence, not decoration. Leading Workshop 2 with Nathalie Gaudreault of the Allen Institute for Cell Science, she made the case for raw data and metadata as a first line of defense against generative AI.

Jana Christopher, FEBS Press

Achal Agrawal wants the people who find errors and the people who fix them in the same room. His Sleuthbox aims to open a direct, trackable channel between sleuths and publishers.

Achal Agrawal, Founder of India Research Watch

Mike Perkins pressed a single question across two sessions: how can institutions make AI use transparent and accountable without reducing integrity to a compliance exercise? His answer, whether in teaching, assessment or the emerging Vancouver Standard for research, was that principles are not enough.

Mike Perkins, British University of Vietnam

Brenda Adhiambo Odero used her winning presentation at Oral Session 1G to point at what disclosure forms leave out: the people who build the tools. Kenyan workers, she noted, labeled toxic content for as little as $1.32 an hour to make AI safe to use, then rated millions of outputs to shape the writing style integrity checks now flag as "AI-generated."

Brenda Adhiambo Odero, University of KwaZulu-Natal

Nihar Shah delivered his talk on autonomous AI scientists as the "Human Scientist Memorial Lecture," dressed as the Grim Reaper and shedding a piece of the costume with each flaw he exposed. By the end, the memorial was postponed until further notice.

Nihar Shah, Carnegie Mellon University

Panagiotis Kavouras brought two years-in-the-making efforts to Vancouver: the Athens Statement on embedding research integrity in cross-sectoral research and policymaking, and early findings from the TRUSTparency project, which is building reproducibility practices for universities, funders and publishers.

Panagiotis Kavouras, University of Oslo

One of our most precious memories of the 9th WCRI in Vancouver was the deliberate engagement of early career researchers and professionals (ECRPs) in the research integrity community. Over the past five years this network has grown into a community of engaged, positive and enthusiastic research integrity professionals and like-minded people looking for support and professional growth in their roles back home. Our goal as organizers was to bring that energy together: to share our passion for research integrity and learn from one another. We hold to the African proverb:

“If you want to go fast, go alone; if you want to go far, go together.”

So what did we do? We opened with an interactive pre-conference workshop on building an inclusive workplace culture that supports research integrity. Participants spoke openly about experiences with leadership and the tools they rely on to create supportive environments. Then, we held an informal networking session, bringing ECRPs together with veterans of the field, and offered a safe space to talk openly, get to know each other and exchange ideas. ECRP supporters from the first hour joined us in person, including Lex Bouter and Robbert Dijkgraaf. They showed us that "early career" is above all a mindset, with far less to do with age.

Many participants valued this more intimate setting, which made room for the deeper conversations and relationship-building that large international conferences rarely allow. Together, these moments showed the value of dedicated spaces where ECRPs can connect, contribute and feel they belong to our vibrant and passionate community.

We flew home with backpacks full of stories, ideas and connections. Two thoughts kept resonating. First, we were struck by how engaged and passionate our early career community is, and how strongly we felt that spirit among us. That left us grateful and full-hearted. Second, a little nostalgia: it will be two years before we meet in person again, in Tallinn at the next WCRI in 2028.

Paula Saner, Joeri Tijdink and De-Ming Chau

Coordinators, ECRP Network

Conference Awardees

Excellence in Doctoral Research

Joonha Jeon for his work on the emerging market of research integrity detection tools, which has expanded with the rise of paper mills; and Emmanuelle Piedboeuf for a project examining how Indigenous organizations in Quebec are reshaping the governance, ethics and practice of research according to their own priorities and IKS.

Best Poster

• Catriona Leslie

• Ruben van Diest

• Amani Abu Shaheen

   

Best Oral Presentation

• Brenda Odero

• Marijke van Vlasselaer

• Ilse de Jong

Anderson Kleinert Award

• Allen Mukhwana

• Vladyslava Kachkovska

• Evan Laye and Lee Dixon (joint)

Catriona Leslie, Best Poster Award

Marijke van Vlasselaer, Second Prize, Best Oral Presentation

Joonha Jeon, Excellence in Doctoral Research Award

As co-chairs of WCRI 2026, we are deeply encouraged by the breadth of engagement and the collective commitment to research integrity shown throughout this conference. We are also grateful to SIA for dedicating space in REACH to the important themes and discussions emerging from Vancouver, and for helping extend these conversations to a broader community.

The discussions at WCRI 2026 have reinforced that research integrity is not a static goal but an ongoing, shared responsibility that spans disciplines, interest holders and borders. We leave the conference with renewed momentum, grounded in evidence, collaboration and a clear recognition that strengthening research integrity requires both local action and global coordination.

We look forward to seeing how these conversations translate into sustained action across the research ecosystem.

David Moher, Lex Bouter and Chris Graf

Co-Chairs, WCRI 2026

From Conversation to Action

The conference co-chairs said it perfectly: research integrity is a shared responsibility that does not pause between conferences. WCRI takes place every two years, but the conversations, collaborations and practical work continue every day in between.

Turning that momentum into action is at the heart of SIA's mission.

SIA brings together researchers, research integrity specialists, organizations and stakeholders committed to strengthening research integrity around the world. We connect people across disciplines, sectors and regions, creating opportunities to share knowledge, learn from one another and develop practical solutions to shared challenges. By bringing diverse perspectives together and combining the impact of those working in this space, we aim to help build a more transparent, trustworthy and collaborative research ecosystem.

SIA's interconnected platforms are designed to support that work. REACH shares timely analysis and perspectives from across the research community. Together with our partners, we also continue to develop resources that offer evolving guidance, education, collaboration and opportunities for knowledge exchange.

If you share our commitment to strengthening research integrity, we invite you to support our work by subscribing. You'll stay connected with a growing international community, receive expert insights and practical resources, and contribute to shaping future discussions, collaborations and initiatives. Together, we can keep the momentum generated at WCRI going long after the conference has ended.

The conversations from Vancouver are already carrying on in the HIKE forum, where subscribers can read exclusive interviews with Ry Moran, Retha Visagie, Robbert Dijkgraaf and Bert Seghers, and engage with SIA's research integrity network.