AI system transparency registers

A user has no way to establish which system is handling an interaction, who built it, where it is hosted, or what data it was trained on, and neither does the agent acting on their behalf. That gap widens as more government services come to run on AI. Without that account, a decision cannot be checked against who is accountable for it, and the jurisdiction the system answers to stays invisible at the moment it matters.

The system behind a service has to be identifiable to a user or an agent at the point of use, not only to those who go looking in a register.

The UK Algorithmic Transparency Recording Standard. The UK mandated ATRS reporting across all government departments, and the scope policy extended it to arm's-length bodies providing public or frontline services or routinely interacting with the general public. What is caught are tools that significantly influence a decision with public effect, or interact with the public directly, and the standard requires disclosure of purpose, training data, underlying technology, and risk-management strategies. More than 125 records have been published.

The Netherlands Algorithm Register. The Netherlands launched a public algorithm register inviting government bodies to publish the algorithms they use rather than obliging them to, and 1,520 are listed. The Dutch Data Protection Authority found that more than half of municipalities and over three-quarters of independent administrative bodies had registered nothing, that only 5 percent of registered algorithms carried a fundamental-rights assessment, and called for registration to be made mandatory. The distance between what is listed and what is in use is the measure of a voluntary register.

EU AI Act Article 50, transparency obligations. Providers of AI systems intended to interact directly with natural persons must ensure those persons are informed they are interacting with an AI system, and providers of systems generating synthetic content must mark outputs in a machine-readable format. Non-compliance draws administrative fines of up to EUR 15 million or 3 percent of worldwide annual turnover. The higher tier applies to prohibited practices, and not to transparency breaches.

EU AI Act Article 13, information for deployers. Providers of high-risk AI systems must deliver instructions that are 'concise, complete, correct and clear', covering intended purpose, accuracy metrics, and details of training and validation data. The duty creates an information chain running provider to deployer to user.

The ATRS and Dutch register models provide the structural template for AI system disclosure. However, they are currently registry-based, published on a government website for those who seek them out. In an agent-mediated interaction, the disclosure needs to be surfaced at point of use, not discovered in a register after the fact.

The design challenge is adapting registry information into a glanceable, contextual indicator. The Article 50 obligation provides the legal mandate and the registries provide the data; the part still to be built is the interaction that brings that data to the user at the point of use.

A government can already disclose which model powers a function at the level of a procurement notice, as France did in notifying its framework agreement with Mistral AI. What does not follow from that is the disclosure a user meets: a procurement notice is read by those who go looking for it, which is the same limitation the registers carry.