Who will control the means of discovery?

Who will control the means of discovery?

20/11/2026
AI is beginning to accelerate the research that produces better AI. Europe needs to connect scientific ambition with independent scrutiny, fair access and a public say in who benefits.

In this edition

  • Editorial: Who will control the means of discovery?
  • Brussels Policy Watch: the Horizon Europe budget and the EU Inc. proposal
  • Inside INESC: patents, networks and dual-use testing
  • INESC Brussels HUB in Action: EARTO, AI-SECRETT, our new website and RIFF
  • Dates worth acting on

Editorial: Who will control the means of discovery?

Europe could become much better at using AI while becoming more dependent on decisions taken inside a handful of companies. For the research community, that dependency could reach beyond the tools we buy to the pace and direction of discovery itself. This deserves a more central place in our discussion of competitiveness and scientific sovereignty.

The METR assessment published on 5 November brings this question into sharper focus. Examining Anthropic’s internal evidence, the independent evaluator estimates that progress on a capability index covering well-specified tasks accelerated to 1.9 times the rate observed between March 2024 and October 2025. It judges that AI-assisted R&D probably accounts for most of the current acceleration, while emphasising substantial uncertainty. This is a consequential finding about one laboratory, rather than a measurement of scientific productivity across the economy.

Anthropic’s own measurements help explain the mechanism. By August, it classified AI as leading 26% of its measured AI R&D work, completing most of a task under human supervision. None of the measured work was classified as fully autonomous. The prospect demanding attention is a feedback loop: better AI helps researchers build better AI.

In his 13 November essay, Dwarkesh Patel argues that automated researchers could also learn to extract more knowledge from small, inexpensive experiments, improving algorithms and the use of computing resources. His forecast of superintelligence is speculative. The mechanism warrants attention independently of that forecast: a technology that improves the process through which its successors are developed can compress the time available for institutional adaptation.

There are good reasons to resist extrapolating this into every laboratory. Physical experiments, reliable data, energy and scientific validation impose constraints that software benchmarks cannot capture. Accelerating one stage of research may simply move the bottleneck elsewhere. Yet uncertainty about the eventual pace leaves an immediate political question: who will have the capacity to investigate what is happening, and act on what they find?

METR reports that Anthropic redacted most of the evidence it found most informative. Commercial confidentiality can be legitimate; the resulting information imbalance still matters. In a separate review, METR considered it plausible that the company’s early-warning threshold had already been crossed, while agreeing that direct catastrophic risks from the July models remained low. These are exactly the judgements public institutions need the expertise and access to scrutinise.

Europe has foundations on which to build. The AI Office’s October report on independent evaluators addresses their qualifications and independence. The next political test is whether independent evaluation has durable funding, meaningful access and the ability to communicate findings in the public interest, with appropriate safeguards for sensitive information. Governments need an understanding of the frontier that they can interrogate themselves.

The same question of public capacity runs through Europe’s scientific ambitions. The AI in Science Strategy establishes RAISE as a means of pooling talent, data, computing resources and research funding. It also sets out a €600 million Horizon Europe contribution to secure access to AI Gigafactories for researchers and startups. These initiatives could strengthen the ability of public science to participate in an accelerating research process.

Their success should therefore be judged partly by the independence they make possible. Can researchers examine and reproduce results when a model changes? Can institutions move between suppliers without losing their research workflows? Can a smaller institute obtain sustained access on workable terms? If the institutions with the best AI gain an advantage in developing the next generation of research tools, unequal access could compound. Scientific sovereignty will depend on the conditions under which researchers can work.

It will also depend on how they learn. CESAER’s position of 10 November argues for strong disciplinary foundations and independent judgement, backed by investment in teaching and institutional capacity. This raises an uncomfortable question for research organisations: if we automate the tasks through which junior researchers acquire experience, how will we develop the people expected to supervise the systems? Invoking “human oversight” does little unless we sustain the expertise that makes oversight meaningful.

There is a deeper scientific concern here. Lisa Messeri and Molly Crockett’s analysis in Nature warns that AI can encourage an illusion of understanding and narrow the range of questions and approaches pursued. Research policy should take seriously the possibility that greater output and greater understanding may diverge. Protecting methodological diversity and the freedom to pursue difficult, unfashionable questions becomes particularly important when funding and evaluation reward speed.

The consequences extend beyond research. If AI substantially raises the productivity of cognitive work, decisions about ownership, competition and taxation will help determine who benefits. Workers’ bargaining power, access to retraining and the financing of public services belong in this discussion now. Predictions of imminent mass unemployment go beyond what the METR study establishes. Governments can nevertheless prepare for different outcomes and identify which developments should trigger a response. Preparedness requires decisions under uncertainty; it cannot depend on a settled forecast.

As FP10 and the European Competitiveness Fund take shape, these questions should influence the terms of public investment. Support for AI development and deployment should help build shared capabilities, enable independent scrutiny and widen access to the gains. The public return should be visible in access to knowledge and better services, alongside commercial success. Long-term funding remains essential. Alongside it, funders need ways to revise methods and redirect resources as evidence changes, without requiring researchers to anticipate every technical development years in advance.

For the INESC community, this is also an institutional responsibility. Across AI research, engineering and industrial applications, we can test where these systems improve reliable outcomes, where they shift costs onto verification, and what smaller firms need to benefit. Our contribution to policy will be stronger if it draws on evidence about changes in research practice, including the skills and infrastructure those changes require.

The Commission and Member States should bring these questions into a common preparedness agenda, linking RAISE and independent evaluation with research on the economic and social consequences of AI. That requires funded work on plausible scenarios, public reporting and explicit responsibility for acting when the evidence changes. Researchers, employers and trade unions should help shape it.

Faster discovery could help Europe address problems it has struggled with for decades. Whether it does so will depend on choices about research priorities, access and the distribution of benefits. The discussion we need is about our collective ability to make those choices while they are still ours to make.


Brussels Policy Watch: Horizon Europe budget and the EU Inc. proposal

Will Parliament’s €200bn hold against the Council’s €140bn?

On 10 November, ITRE backed Parliament’s Horizon Europe position by 67 votes to 1, with 18 abstentions. That position asks for €200bn, the figure Draghi named for the programme. The Council has offered €140bn, which leaves a €37bn gap once both are compared at 2025 prices. Rapporteurs Ehler and Repasi have taken the case to Berlin, and on 9 November 29 European research organisations issued a joint statement and petition calling on European leaders to commit at least €200bn to Horizon Europe 2028–2034 and to invest in research, development and innovation for a stronger Europe tomorrow.

Sources: Joint statement: Fund RD&I, Fund the Future · Science|Business

The 28th regime is turning into 27 national ones

The Council and Parliament are both rewriting EU Inc. (also known as the 28th regime), the Commission’s March proposal for a single EU-wide company form for start-ups and scale-ups. In their 23 October mandate agreed by EU ambassadors, Member States removed the EU insolvency procedure. They also restricted the new EU employee stock options, taxed only when shares are sold, to companies with up to 750 employees, so scale-ups lose them as they grow. And they gave themselves 21 months instead of 12 to put the new company form in place. Parliament’s draft would add national labour and co-determination safeguards. Legal scholars warn the result could be “27 regimes wearing the same badge.”

Parliament’s committee vote has been postponed repeatedly over workers’ rights, and without a Parliament position, negotiations with the Council cannot start. A 2026 deal now looks unlikely. Some voices now argue that the proposal should be withdrawn rather than adopted in a weakened form. EIC Board member Ann Mettler called on the Commission on 10 November to pull what she described as a “completely watered down” text, and the same legal scholars made the same call that day.

Sources: Science|Business · ECGI · Eunews · European Commission


Inside INESC: From research results to European reach

Recent developments across the INESC system show different ways in which scientific and technological capabilities create value beyond individual projects: through intellectual property, European collaboration, international research communities, cross-regional innovation, real-world validation and new applications of emerging technologies.

INESC TEC leads Portuguese research organisations in European patent applications

A new European Patent Office study places INESC TEC first among Portuguese organisations for European patent applications, with 14 applications filed, five jointly owned with the University of Porto.

The technologies span areas including photonics and advanced sensing, digital health, clean energy and smart grids, robotics, computer vision and digital twins. Beyond the ranking itself, the figures point to something more relevant for INESC: the role of intellectual property in helping research move towards industrial development, investment and commercialisation.

Source: INESC TEC

INESC INOV-Lab brings a new international communications community to Lisbon

INESC INOV-Lab and Instituto Superior Técnico are co-organising IEEE ISAC 2026 in Lisbon from 15 to 18 November, the first major international IEEE conference dedicated specifically to Integrated Sensing and Communications.

The field combines communications and sensing within the same systems and is increasingly relevant to the development of future wireless technologies. The programme brings together researchers and industry experts around scientific advances, standardisation and the future development of ISAC technologies.

For INESC INOV-Lab, co-organising the conference provides more than visibility: it places the institute within an emerging international community at the point where research, future standards and industrial applications are beginning to converge.

Source: INESC INOV-Lab

A new network connects European and Latin American HPC and AI ecosystems

INESC TEC is part of the EU-LAC Supercomputing Network for AI, a two-year initiative connecting the European Union with Latin America and the Caribbean around high-performance computing (HPC) and artificial intelligence.

The network focuses initially on Argentina, Brazil, Chile, Colombia, Costa Rica, Mexico and Uruguay. Its work combines policy dialogue, training and skills, joint use cases and computing connectivity, responding to challenges such as fragmented HPC capacity, unequal access to resources and shortages of specialised skills. Planned applications include climate modelling, energy efficiency, natural-hazard prediction, health and large language models.

The initiative is particularly relevant because it treats computing capacity, expertise and international connectivity as parts of the same ecosystem rather than separate issues. It also builds on the BELLA research and education connectivity infrastructure linking Europe and Latin America through the EllaLink submarine cable, whose main European landing station is in Sines.

Source: INESC TEC

Dual-use robotics moves from the laboratory into operational testing

INESC TEC took its Modular-X robotic platform and a modular solar-charging solution into ARTEx26, the Portuguese Army’s annual technological experimentation exercise, which brought together 70 organisations from 22 countries.

The tests placed technologies already used in civilian environments such as forestry and industry into demanding defence scenarios. They confirmed the platform’s mobility and robustness, while also revealing requirements that become more important in operational environments, including speed of execution, direct teleoperation and redundant communications. The solar-charging solution was tested as a way of increasing robotic autonomy in remote environments with limited access to conventional charging infrastructure.

The value of the exercise lies in that transition from technology development to validation with end users. It shows how mature capabilities can be adapted to different operating requirements, and how civilian technologies can acquire a credible dual-use pathway through testing rather than through relabelling alone.

Source: INESC TEC

INESC-ID brings generative AI into European cultural heritage

INESC-ID has joined NARRATIA, a Horizon Europe project combining generative AI and mixed reality to create personalised experiences at cultural heritage sites. Coordinated by the Universidad Politécnica de Madrid, the €4.3 million project brings together 13 partners from seven countries and will test the technology across six European heritage sites. INESC-ID will contribute expertise in AI-generated interaction and game-based learning, illustrating how Portuguese research capabilities can connect emerging AI technologies with European applications beyond traditional technology sectors.

Source: INESC-ID

INESC Coimbra brings smart technologies and energy expertise to COP31

Prof. Carlos Henggeler Antunes, Director of INESC Coimbra, joined the University of Coimbra’s official mission to COP31 in Antalya, contributing to discussions on how scientific knowledge can move more effectively into climate policy and implementation.

At the Higher Education for Climate Action Pavilion, he took part in the roundtable “From Strategy to Implementation: Accelerating the Climate Transition through Science, Governance and Territories”, which brought together perspectives on how universities, governments, businesses and citizens can translate knowledge into concrete climate action.

His contribution focused on the role that smart technologies and data can play in the sustainable management of energy and resources.

For INESC, the relevance goes beyond representation at COP31. It shows how expertise in energy systems, optimisation, data and decision support can contribute directly to one of Europe’s and the world’s central challenges: turning climate ambition into policies and actions that can actually be implemented.

Source: University of Coimbra


INESC Brussels HUB in Action: Policy dialogue, projects and a new website

INESC contributes to EARTO discussion on the future governance of Technology Infrastructures

Ricardo Miguéis represented the INESC Brussels HUB in EARTO’s webinar on the future EU governance of Technology Infrastructures, alongside speakers from RISE, Helmholtz/DLR, CEA and VTT.

The discussion comes as Europe considers how Technology Infrastructures should be coordinated at EU level and how operators, industry and public authorities should be represented in the emerging governance framework.

For the HUB, participation provides a direct connection to a European policy discussion that could shape how technology capabilities are coordinated, accessed and supported in the next funding period.

Source: EARTO webinar recording

AI-SECRETT moves from consortium planning to sector-level discussion

The AI-SECRETT consortium met in Las Palmas de Gran Canaria from 19 to 21 October for its third Technical and General Assembly Meeting, bringing partners together to review progress and align the project’s next phase.

Discussions covered the Cascade Funding Call, the AI-SECRETT Master’s programme, the Content Delivery Platform, the AI-STEAM Network, communication activities and recent European policy developments around AI and skills.

That work continued through European AI Innovation Month, which concluded for AI-SECRETT on 10 November with the Transport and Mobility webinar. The INESC Brussels HUB contributed to this final session, connecting the project’s work on skills with wider questions around AI adoption, workforce needs and the future of mobility.

Source: AI-SECRETT

INESC Brussels HUB’s renewed website: a new digital front door for the INESC ecosystem

The INESC Brussels HUB has launched its newly revamped website, with a clearer presentation of the HUB’s role, the five INESC institutes and their connection to the European research and innovation ecosystem.

The new site also includes a Research and technology infrastructures catalogue, bringing together research capabilities and infrastructures from across the INESC ecosystem in one place.

More than a visual update, the website strengthens the HUB’s role as a European-facing entry point into INESC, making it easier for partners, institutions and other stakeholders to understand the breadth of expertise available across the five institutes and identify relevant capabilities for future collaboration.

Source: INESC Brussels HUB

RIFF brings Ukraine’s research infrastructure recovery into the Falling Walls debate

RIFF was represented at this year’s Falling Walls Science Summit in Berlin, where Ricardo Miguéis took part in the non-public Executive Table “Rebuilding Science in Ukraine” alongside RIFF partners from Kyiv Academic University, Jagiellonian University/SOLARIS and DESY.

The discussion explored how the recovery and closer European integration of Ukraine’s research infrastructures can support the rebuilding of its scientific system. Ricardo’s contribution focused on research and technology infrastructures as instruments of science diplomacy, and on the steps needed to move from common evidence and coordination towards deeper integration, commitment and co-ownership with Ukrainian partners.

The exchange also provided an opportunity to connect RIFF’s current work (including the development of a shared evidence base and a national roadmap for Ukrainian research infrastructures) with wider European discussions on Ukraine’s long-term integration into the European Research Area.

Source: RIFF project


Dates worth acting on

19 November · Empowering Europe with AI Skills: AISHA, AI4Gov-X & AI-SECRETT Joint Event Brussels and online. Brings together the European Commission, regional leaders, industry and three EU-funded skills projects to examine how AI competence can move from basic literacy towards trusted and recognised skills. The discussion will cover micro-credentials, sector-specific training and the future AI Skills Academy. Relevant to teams working on AI, skills, public-sector transformation and workforce development. Registration open.


23 November · Building blocks for future economic analysis and modelling: the role of AI Seville and online. Brings together AI experts, economists, researchers and policymakers to examine how AI can support economic analysis and modelling across areas including climate, energy, transport, investment and innovation policy. Relevant to teams working on AI, modelling, optimisation and evidence-based policy.


25 November · AU-EU-Med Innovation Fair on Food and Nutrition Security & Sustainable Agriculture Valenzano, Italy. Brings African, European and Mediterranean R&I actors together around food systems, agriculture and innovation cooperation. Relevant to teams working on agri-food, sustainability and international R&I partnerships.


26 November · Horizon Europe Cluster 1 Info Day Online. The European Commission will present the 2027 Health work programme and forthcoming funding opportunities. Relevant to INESC teams working across health, biomedical AI, data and digital technologies.


2 December · European Life Sciences: Driving an Innovative Economy Brussels and online. Commission conference reviewing implementation of the European Life Sciences Strategy, with sessions on AI, biotechnology, food systems, investment priorities and research ecosystems.


2 December · CEF Digital Community Conference 2026 Brussels. Focuses on Europe’s future digital connectivity and forthcoming investment priorities, including links with the European Competitiveness Fund. Relevant to communications, digital infrastructure and connectivity expertise.


2–4 December · International Conference on Research Infrastructures (ICRI 2026) Major international meeting bringing together policymakers, funders, operators and researchers around the future of research infrastructures and international cooperation. Relevant to FUND-RI, RIFF and wider INESC infrastructure work.


3 December · Choose Europe 2026 – Advancing Research, Innovation and Talent Munich. High-level Commission and German Government conference on strengthening Europe as a destination for scientific, innovation and education talent. Relevant to research careers, talent attraction and institutional capacity.


 

4 December · 1+ Million Genomes Initiative – High-Level Event Online streaming. Focuses on Europe’s progress towards secure cross-border access to genomic and health data for research, healthcare and policy. Relevant to health data, AI and personalised medicine capabilities.

 


 

8 December · EU Knowledge Valorisation Talks Online. Commission series on practical approaches to improving the uptake and use of research results. Programme details are still to be announced. Relevant to technology-transfer, valorisation and innovation teams across INESC.

The Insider brings together selected developments from European research, innovation and policy, alongside updates from across the INESC ecosystem. Find out more about the INESC Brussels HUB at hub.inesc.pt.

Tags: Artificial intelligence; AI in science; AI governance; EU research and innovation policy; Research careers; Scientific sovereignty

 

STAY CONNECTED

The Insider Newsletter brings together selected developments from European research, innovation and policy, together with updates from the INESC ecosystem.

For further information, visit the INESC Brussels HUB website or follow us on LinkedIn.

Related policy

Nothing linked yet.

Related events

Nothing linked yet.

Related news

Nothing linked yet.