AI Act Article 53(1)(d) Mandated Transparency of Training Content: Evidence of Feasibility and Widespread Non-Compliance

Article 53(1)(d) is the AI Act's only publicly accessible transparency measure that addresses known issues regarding impact of training on fundamental rights. It requires General Purpose AI (GPAI) model providers to publish a detailed public summary based on the template provided by the AI Office. To support this objective, we undertook a methodological discovery process and found 64 summaries, which we scored for transparency (i.e. if required information was provided) and usefulness (i.e. if it is actionable). We found a few showing excellent scores, though most scored poorly across both dimensions. We also found many models, often by these same providers, were missing summaries despite this being an enforceable legal obligation. Our analysis also identifies critical issues such as difficulties in discovering summaries despite a legal obligation of disclosure alongside distribution channels, as well as missing or inadequate information, such as regarding sources and methods used for web scraping, that limit the ability to identify potential rights violations. Given that the phased enforcement has already begun from 2nd August 2026, our work is extremely timely as it empowers legal enforcement and rights stakeholders, and supports the AI Act by demonstrating quantitatively (1) how public summaries can be an effective instrument for facilitating information disclosure, and that (2) some providers are intentionally seeking to avoid accountability. To support future actions, we identify the most pressing issues for each section of the summary and compare scores across strategic groups based on open/closed weights, Code of Practice signees, and jurisdictions. We have also disseminated our findings with key stakeholders including the EU Commission's AI Office, EU Parliament, civil society, and rights holders.

Authors

Publication Details

Journal
OSF Preprints (OSF Preprints)
Published
2026-11-17
DOI
https://doi.org/10.31219/osf.io/v5tma
Primary Topic
Digitalization, Law, and Regulation
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

AI Act Article 53(1)(d) Mandated Transparency of Training Content: Evidence of Feasibility and Widespread Non-Compliance

Harshvardhan Jitendra Pandit, Dick A. H. Blankvoort
4 links
OSF Preprints (OSF Preprints)
Digitalization, Law, and Regulation
preprint

AI Act Article 53(1)(d) Mandated Transparency of Training Content: Evidence of Feasibility and Widespread Non-Compliance

Harshvardhan Jitendra Pandit, Dick A. H. Blankvoort
preprint en

Abstract

Article 53(1)(d) is the AI Act's only publicly accessible transparency measure that addresses known issues regarding impact of training on fundamental rights. It requires General Purpose AI (GPAI) model providers to publish a detailed public summary based on the template provided by the AI Office. To support this objective, we undertook a methodological discovery process and found 64 summaries, which we scored for transparency (i.e. if required information was provided) and usefulness (i.e. if it is actionable). We found a few showing excellent scores, though most scored poorly across both dimensions. We also found many models, often by these same providers, were missing summaries despite this being an enforceable legal obligation. Our analysis also identifies critical issues such as difficulties in discovering summaries despite a legal obligation of disclosure alongside distribution channels, as well as missing or inadequate information, such as regarding sources and methods used for web scraping, that limit the ability to identify potential rights violations. Given that the phased enforcement has already begun from 2nd August 2026, our work is extremely timely as it empowers legal enforcement and rights stakeholders, and supports the AI Act by demonstrating quantitatively (1) how public summaries can be an effective instrument for facilitating information disclosure, and that (2) some providers are intentionally seeking to avoid accountability. To support future actions, we identify the most pressing issues for each section of the summary and compare scores across strategic groups based on open/closed weights, Code of Practice signees, and jurisdictions. We have also disseminated our findings with key stakeholders including the EU Commission's AI Office, EU Parliament, civil society, and rights holders.

OSF Preprints (OSF Preprints)
Peace, justice, and strong institutions
Digitalization, Law, and Regulation
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon