Topic outline

  • Ethical Challenges in the ENVRI-Hub design: Data Reuse, Aggregation and Sustainable Storage

    Chapters 
    • 00:00 Introduction
    • 07:24 Harmonising Open-Science Principles with Data Sovereignty and the Mitigation of Ecological and Security Threats
    • 38:58 Ethics and Data Aggregation
    • 01:01:09 Is my Data Prone to Dual Use?
    • 01:19:24 Questions and answers

    • Why this training session

      This ENVRI-Hub NEXT training webinar was organised on 3 September 2026 as follow-up of the first session: "Ethical Challenges in the ENVRI-Hub design: AI, Data Stewardship and User Rights".

      In this second webinar, our ethics experts clarify the theoretical framework around data sovereignty, unveil the tension between open science and dual-use research, and suggest concrete measures for ethical data management.

      • Presentations

        • Harmonising Open-Science Principles with Data Sovereignty and the Mitigation of Ecological and Security Threats File

          In her presentation, Professor Tsinorema outlines key ethical concepts and challenges related to data sovereignty and sustainable openness.

          Data Sovereignty vs. Data Ownership

          "We are our data. We don't possess our data: the violation of personal data is a violation of personality rights. That's why I don't want to talk about ownership, but sovereignty."

          Data control should be viewed through the lens of "sovereignty", i.e., legitimate political authority and control, rather than mere "ownership," as personal data inherently reflects human personality rights.

          Harmonising Data Sovereignty and Openness

          While open science is a core scientific value vital for reproducibility and the public good, unconstrained access is not automatically ethical. Openness must be harmonised, rather than strictly balanced, with data sovereignty and security. Responsible stewardship entails harmonising societal benefits stemming from collaboration and open science with legitimate protection of data subjects, communities and institutions against unauthorised second use.

          Challenges
          • Limitations of traditional consent: In the context of AI, big data, and interoperable datasets, standard individual informed consent models and traditional anonymisation/de-anonymisation polarities are no longer normatively sufficient to protect human subjects. Datasets may be harmless, but when combined, they can expose vital information in unpredictable ways;
          • AI and algorithmic concerns: The increasing use of AI in data processing raises significant ethical concerns, particularly regarding algorithmic bias when utilising datasets that lack diversity or high quality;
          • Ecological risks: Energy-consuming digital infrastructures create environmental costs. Such costs for data use and preservation must be always clearly justified by scientific and societal value;
          • Security tensions: Unrestricted open data carries inherent security risks, including malicious or illegitimate dual-use, harmful cyber attacks, and exposure of geopolitical vulnerabilities.
          Data Sovereignty by Design and Sustainable Openness
          The traditional standard of "as open as possible, as closed as necessary" should be refined to: "as open as possible, restricted as ethically required, and sustainable as reasonably possible," relying on human ethical judgment and graduated access.

          Ecological sustainability, security, and data sovereignty are not merely external restrictions added onto research, but core intrinsic components of conducting open science responsibly. Success relies on developing a robust governance framework that integrates strong data stewardship with ethical standards, which are essential for the quality of scientific work.

          How much openness is ethically justified? Which types of data should be accessible to whom, for what purposes, and under which authority? Answering these questions requires the scientific community to engage in continuous dialogue and re-evaluation of ethics assessment tools and guidelines.

        • Ethics and Data Aggregation File

          Professor Louise Bezuidenhout shifts focus to the practical and systemic ethical risks introduced by data aggregation, namely, the recombination of multiple datasets within interconnected infrastructures and research commons like EOSC. While data aggregation yields valuable insights impossible with isolated datasets, it creates complex issues regarding provenance, power dynamics, and responsibility that extend beyond traditional privacy concern.

          Ethical Challenges in Data Aggregation
          1. Privacy and re-identification: Combining de-identified, seemingly harmless datasets can expose sensitive variables (such as location, age, or demographics) and re-identify individuals or groups.
          2. Unpredictable secondary use: Aggregation often repurposes data beyond its initial collection scope without feedback mechanisms to inform original contributors. Ethical evaluation must shift from asking "Is this data valuable for society?" to "Does the research align with the original intentions and commitments made to contributors?".
          3. Loss of individual and community agency: Merging datasets makes it difficult for contributing communities to track, correct, or withdraw their data. This is particularly critical for indigenous communities under the CARE principles, where the authority to control data must be respected.
          4. Hidden biases: Aggregation can obscure embedded biases rather than remove them. The paradigm must shift to evaluating who gets represented by the aggregated dataset and who becomes invisible Infrastructural
          5. Geopolitics vulnerabilities: Federated structures bring risks related to geopolitical instability, malicious cyberattacks, and sudden funding disruptions. Metadata, documentation, and governance frameworks carry explicit ethical weight in protecting data sovereignty and maintaining custody chains.
          6. Power dynamics and the "FAIR fallacy": Data choices are not value-neutral; those who possess the power to combine datasets decide which worldviews and categories are normalised. Technical interoperability (FAIR principles) does not automatically guarantee epistemic comparability or equitable representation.
          7. Unsustainable openness: Recognising the environmental costs of energy-consuming storage, research must move away from indiscriminately saving all data. Instead, projects should adopt "sustainable openness" by strategically determining what data requires immediate accessibility, what can be moved to long-term dark vaults, and what metadata suffices.
          Institutional Recommendations and Practical Measures

          Professor Bezuidenhout highlights a "cultural lag" where technical data capabilities outpace social and ethical governance, compounded by an institutional disconnect between Research Ethics Committees (RECs) and technical Data Stewards. To address these challenges, she recommends:

          • Individual actions
            • Cross-disciplinary awareness: Acknowledging that datasets from different disciplines carry distinct requirements and ethical constraints.
            • Critical categorisation: Reflecting on personal worldviews and assumptions when defining categories for aggregated datasets.
          • Institutional measures
            • Strategic collaboration in data management: Actively linking Research Ethics Committees with Data Stewards to co-evaluate data aggregation workflows.
            • Provenance tracking: Implementing robust provenance tools and chain-of-custody documentation to trace data sources will satisfy both legal and social commitments.
            • Infrastructure vetting: Certifying the security of the federated infrastructure (e.g., CoreTrustSeal) to protect members’ data sovereignty.
        • Is my data prone to dual use? File

          Dr. Lihavainen takes a pragmatic approach based on his experience as the Director of the Svalbard Integrated Arctic Earth Observing System (SIOS), answering the question: "Is my data prone to dual use?" with a definitive "Yes."

          Malicious actors are highly capable of combining seemingly innocuous scientific data to extract sensitive intelligence. All digital data is accessible to sophisticated actors, making absolute prevention impossible, though mitigation is achievable.

          Definition of Dual-Use Research and Examples from the Arctic

          "Dual use" refers to the potential for new technology or data originally gathered for beneficial research to be misused for malicious or harmful intent.


          • Polar bears tracking: GPS trackers placed on polar bears by the Norwegian Polar Institute for research and community safety cannot be published in real-time or at high resolutions, despite interest by wildlife tourism. Doing so would expose the animals to harassment or worse. Data is subjected to an embargo period and spatial resolution blurring before release.
          • Conspiracy theories: The High-frequency Active Auroral Research Program (HAARP) antennas in Svalbard, designed for low-energy ionosphere studies, have been subjected to data misuse by conspiracy theorists who falsely claim the infrastructure is a weather modification tool or a dangerous energy weapon.
          • The "Pristine" myth: While whaling has permanently altered the Arctic ecosystem, tourism data often still misrepresents it as untouched.
          • Oceanographic surveillance: Research moorings equipped with hydrophones intended for real-time environmental monitoring were restricted by the Norwegian government because they could easily be dual-used to track marine traffic and submarines. Consequently, data can only be gathered manually once a year.
          • Geopolitical intelligence: Almost all data collected by SIOS, ranging from Arctic sea ice, weather, and seabed information to permafrost and wildlife, has dual-use potential for military planning, infrastructure vulnerability assessment, and commercial intelligence.
          Mitigation Strategies

          While preventing dual use entirely is impossible, Dr. Lihavainen suggests several ways to make it significantly harder:

          1. Risk assessments: Evaluating potential misuse prior to data publication.
          2. Embargo rules: Lowering the spatial or temporal resolution of published data and implementing data publication delays
          3. Authentication and access controls: Following the principle of "as open as possible, as closed as necessary," although establishing who gets to define this boundary requires careful and inclusive considerations;
          4. Technical safeguards: Adopting data passports, encryption, watermarks, and adhering to strict international legislation like GDPR.

          Scientific conferences are increasingly dominated by discussions on research security and AI, since determining exactly who bears the responsibility for harmonising open data with security risks remains a complex, unresolved issue in academia.

      • Previous training session