Transparency by default
Telling the user they are dealing with AI, what it can do, and what it is doing right now, at the moments that matter rather than in a terms page. The standing record lets an auditor confirm, after the fact, that the user was told.
The impact of agents
Users cannot calibrate trust in what they cannot see. When an AI agent acts on a government service, users need to know: that they are interacting with AI (not a human), what the agent can and cannot do, what data it accesses, how it reaches decisions, and what recourse exists if something goes wrong.
Disclosure that sits buried in terms and conditions, or arrives as a technical data sheet, does neither job: the user has to meet it before relying on the agent, and has to be able to read it once they do.
What must be verified
Government needs a user to be able to tell that they're dealing with AI, at the time it matters and not after the fact, so they can calibrate trust in what the agent is doing while it does it. The agency or service operator running the agent must trigger that disclosure at the decision point.
Protecting access
A dense transparency notice gets skimmed or abandoned by the people it should protect. A disclosure gate that demands acknowledgment before service deters them from the service itself.
Keeping the path open
- Make disclosures scannable and skippable without penalty, and never gate access behind acknowledgment.
- Offer more than one format (text, audio, simplified visual), and let the user set a transparency level: brief, standard, detailed.
- Make the moment reachable by keyboard alone and dismissible without losing the user's place in the task.
- Treat transparency as no kind of liability shield: 'we told you it was AI' can never diminish the agency's duty of care to the person who couldn't parse the telling.
Response surface
A standing transparency record is paired with disclosures that appear at the decision points where they can change what the user does.
You’re working with an AI assistant. It helps prepare your energy rebate claim; a human officer decides it. How it works is on record below.
The assistant reads your answers and your household records (with your permission), fills in the claim, and estimates your rebate from the published rate tables. It cannot approve or refuse anything. Officers review every claim it prepares; the assistant’s error and override rates are published quarterly.
Disclosures appear only at decision points and irreversible actions. Raise one at every step and users learn to dismiss it.
The record stays in place, so it can be checked or cited later. A service that publishes only the record hides the moment that mattered; one that shows only moments leaves nothing on record to contest later.
Maturity
- Established
For algorithmic transparency recording, where the UK ATRS is production-grade and mandatory.
- Emerging Headline
For real-time agent transparency, where regulation now demands it but the interaction patterns are still forming.
- Frontier
For adaptive, user-preference-driven disclosure.
Precedents
EU AI Act Article 50, transparency obligations. AI systems interacting directly with people must disclose their AI nature in a clear and distinguishable manner, at the latest at the time of the first interaction. High-risk systems must be sufficiently transparent for deployers to interpret their outputs appropriately, under Article 13. The disclosure is timed to the first interaction, and not left to the interface to place.
UK Algorithmic Transparency Recording Standard. The template runs in two tiers: Tier 1 is a plain-language summary for the public, Tier 2 is technical detail for specialists. It is mandatory for UK central government, with more than 125 records published, and was co-designed with the people who use it. One artifact serves two audiences at different depths.
Australian Standard for AI Transparency Statements. Agencies must publish their approach to AI adoption in a consistent format, and the DTA is planning an addendum addressing agents that initiate actions. The statement is entity-level, and the addendum is the part that would reach an acting agent.
Yocco on necessary transparency moments in agentic AI. Yocco argues transparency should be event-driven, surfaced at decision points, capability boundaries, and error conditions, rather than gathered on a static disclosure page. Placing disclosure at the moment of consequence is what separates it from a page nobody opens.
What carries over to agent use
High transferability for the mechanism. Proactive, comprehensible disclosure of an agent's nature and reach maps onto any government service, and a direct government precedent already exists: the UK's two-tier transparency record (public summary plus technical detail) is immediately adaptable elsewhere. A requirement for upfront AI disclosure at first interaction sets a regulatory baseline that services should meet or exceed.
Key adaptation for agents specifically: static transparency records (ATRS-style) are necessary but insufficient for agents, which act dynamically. Agent transparency requires real-time disclosure: "I am now accessing your tax records to check eligibility" or "I am about to submit this form on your behalf; this action cannot be undone." The "transparency moments" framework provides the design vocabulary for this.
Where things go wrong
Where this goes wrong is an invalid determination method kept out of sight from the people it bills. Mandatory upfront and event-driven disclosure of how a determination was reached exposes flawed logic instead of burying it. The disclosure requirement itself can be met in form only: a notice can satisfy the mandatory-disclosure rule to the letter while its length and jargon tell the person reading it nothing they can act on.
Sources
8 references
The instrument, the operating deployment, or the official record itself.
Writing about the subject rather than the framework itself, including vendor commentary.