Research opportunities and challenges of the EU’s Digital Services Act
Summary
This commentary, authored by researchers actively engaged in EU policy processes, assesses the early implementation of Article 40 of the Digital Services Act (DSA), which obliges very large online platforms (VLOPs) and search engines (VLOSEs) to grant vetted researchers access to internal data. Drawing on an 18-month ERC-supported pilot with DG-CONNECT, national Digital Services Coordinators (DSCs), and roundtables with major platforms, the authors diagnose three interlocking obstacles — misaligned incentives, severe asymmetries between platforms and academia, and operational bottlenecks — and offer three corresponding recommendations: streamline access procedures, fund independent research infrastructure, and close regulatory blind spots, most urgently around large language models. The piece argues that without robust enforcement, the DSA risks being hollowed out by industry resistance and the persistence of “independence by permission” research arrangements.
Key Contributions
- A practitioner-grounded snapshot of Article 40’s operational state as of late 2025, ahead of the first expected data-access decisions in February 2026.
- A tripartite framing of DSA implementation challenges: misaligned incentives, resource asymmetries, and bottlenecks.
- Three concrete policy recommendations: standardized data-sharing frameworks; dedicated funding and community coordination for independent research; and extending DSA scope to LLMs.
- An argument that LLMs (e.g., ChatGPT, Grok) meet DSA criteria for systemic risk as intermediary services and represent a current regulatory blind spot.
Methods
Expert commentary synthesizing the authors’ direct involvement in EU policy deliberations, an ERC-supported pilot program engaging DG-CONNECT and national DSCs, roundtables with platforms and the European Commission, and a review of literature on platform research, AI industry influence, and DSA legal scholarship.
Findings
- Platforms have expressed reluctance toward direct data sharing, occasionally signaling potential legal action against researchers and NGOs.
- Meta’s 2020 US election collaboration illustrates how undisclosed algorithmic changes can bias platform-mediated research outcomes.
- National DSCs and university Data Protection Officers face mismatches in vocabulary and expectations with both researchers and platforms, delaying initial requests.
- A circular resource problem disadvantages early-career and less-resourced researchers: grants require data access, but pursuing access requires funding.
- Industry AI research now dwarfs public research in resources, while DSA-as-censorship narratives have circulated to tens of millions on social media.
- First DSA portal decisions on data-access requests are anticipated in late February 2026.
Connections
This commentary speaks directly to ongoing scholarship critiquing platform-led transparency regimes and the politics of researcher data access — see Rieder2026-pp and Rieder2025-ju on the institutional shape of DSA access, and Helmond2026-ll on platform governance dynamics. The diagnosis of “independence by permission” resonates with concerns about industry-mediated research raised in Bak-Coleman2025-pm and Bak-Coleman2026-mk, while the regulatory blind spot around LLMs connects to broader debates on platform-adjacent infrastructures discussed in Bouchaud2026-lr and Ohme2026-nv.
Podcast
A research-radio episode discusses this paper: Listen