AI4TRUST: What building detection infrastructure tells us about the fight for information integrity

The AI4TRUST project, funded by Horizon Europe, developed a Toolbox and Monitoring Dashboard to detect manipulated content at scale. However, the project encountered significant obstacles in accessing data from Very Large Online Platforms (VLOPs), including rejections from X and Meta. The document argues that DSA Article 40 is failing because platforms unilaterally decide what constitutes 'systemic risk' research. It recommends reforming the DSA vetting process, integrating Foreign Information Manipulation and Interference (FIMI) into risk assessments, mandating infrastructure interoperability, and addressing algorithmic amplification incentives.

• The AI4TRUST project developed a two-part infrastructure for detecting manipulated content: a Toolbox for analyzing individual items (text, image, audio, video) and a Monitoring Dashboard for continuous collection, risk scoring, and human validation.
• Data access for the consortium was fragmented; while YouTube, Telegram, and Bluesky were workable, other platforms blocked access. X rejected a DSA Article 40 application in April 2024, and Meta rejected one in December 2024, claiming the tool would be a prohibited 'derivative work'.
• The implementation of DSA Article 40 is failing, as platforms are effectively deciding what qualifies as systemic risk research. Of 46 applications registered by the DSA 40 Collaboratory, 14 were rejected, often on the grounds that the research did not address systemic risks.
• The document identifies a conflict between public funding for detection tools and platform business models, noting that recommender systems amplify manipulated content for engagement before detection systems can flag it.

APA

Index, G. D. (2026). AI4TRUST: What building detection infrastructure tells us about the fight for information integrity. Global Disinformation Index. https://www.disinformationindex.org/blog/2026-04-20-very-large-online-platforms-decide-what-counts-as-systemic-risk-research-thats-self-regulation-failing/

Chicago

Index, Global Disinformation. AI4TRUST: What building detection infrastructure tells us about the fight for information integrity. Global Disinformation Index, 2026. https://www.disinformationindex.org/blog/2026-04-20-very-large-online-platforms-decide-what-counts-as-systemic-risk-research-thats-self-regulation-failing/.

Wikipedia

{{cite press release |last1=Index |first1=Global Disinformation |title=AI4TRUST: What building detection infrastructure tells us about the fight for information integrity |publisher=Global Disinformation Index |date=20 April 2026 |url=https://www.disinformationindex.org/blog/2026-04-20-very-large-online-platforms-decide-what-counts-as-systemic-risk-research-thats-self-regulation-failing/ |access-date=23 October 2026 |via=Climate Insights Directory}}

BibTeX

@misc{index2026ai4trust, author = {Index, Global Disinformation}, title = {{AI4TRUST: What building detection infrastructure tells us about the fight for information integrity}}, publisher = {Global Disinformation Index}, year = {2026}, month = apr, url = {https://www.disinformationindex.org/blog/2026-04-20-very-large-online-platforms-decide-what-counts-as-systemic-risk-research-thats-self-regulation-failing/}, urldate = {2026-10-23}, note = {Indexed by Climate Insights Directory} }