Skip to main content
Bureau of AI
Category: Content Transparency & Labelling

Disclosure of AI Interaction

Also known as: AI Disclosure, AI Interaction Disclosure, AI Use Disclosure
Simply put

Disclosure of AI interaction is the practice of informing people when they are communicating with an artificial intelligence system rather than a human, or when AI is involved in a decision or output that affects them. The goal is to give individuals enough awareness to interpret the interaction appropriately. In practice, disclosures may cover both direct interactions (such as a chatbot) and less visible uses of AI, such as when AI contributes to a consequential decision.

Formal definition

Disclosure of AI interaction refers to transparency measures intended to reveal the involvement of automated or AI systems in an interaction, process, or outcome, so that affected parties can recognize that they are engaging with, or being influenced by, AI. As commonly framed in the evidence, disclosure typically addresses at minimum two scenarios: (1) notice that a person is interacting directly with an AI system, and (2) notice that AI is used in consequential decisions. Scope and specificity vary across contexts and frameworks—ranging from participatory approaches derived from transparency obligations, to structured schemes such as the Artificial Intelligence Disclosure (AID) Framework for documenting AI use in writing and research, to institutional AI disclosure statements describing how AI was used in an assignment or project. This entry addresses disclosure as a transparency practice and governance control; it does not resolve the operational questions of who must disclose, what must be disclosed, and when, which remain context-dependent and are treated as open design questions in the cited literature. Note also that disclosure is a transparency mechanism and is not established to be uniformly beneficial—at least one cited study reports that AI disclosure can, in some settings, reduce trust. Specific legal disclosure requirements are jurisdiction- and framework-dependent and are out of scope for this definition.

Why it matters

Disclosure of AI interaction addresses a foundational transparency question in AI governance: whether people know when they are engaging with, or being affected by, an automated system rather than a human. As AI becomes embedded in customer-facing tools, decision workflows, and content creation, the absence of disclosure can leave individuals unable to interpret an interaction appropriately or to calibrate how much weight to give an output. Commentary in this area reflects broad agreement that organizations should disclose when customers are interacting with AI and when AI is used in consequential decisions, positioning disclosure as a core element of maintaining trust in AI-mediated processes.

Disclosure is best understood as a transparency mechanism and governance control rather than a guaranteed benefit. The cited literature does not establish that disclosure is uniformly positive; at least one study reports that AI disclosure can, in some settings, reduce trust rather than build it. This tension matters for practitioners because it means disclosure design is not merely a compliance checkbox—the manner, timing, and framing of a disclosure can materially affect how it is received. Treating disclosure as automatically trust-enhancing would overstate what the evidence supports.

The practice also spans a range of contexts, from real-time notice in direct interactions such as chatbots, to structured documentation of AI use in writing and research through schemes like the Artificial Intelligence Disclosure (AID) Framework and institutional AI disclosure statements. Because the operational questions of who must disclose, what must be disclosed, and when remain context-dependent and are treated as open design questions in the cited literature, organizations should approach disclosure as an evolving area rather than a settled set of requirements. Specific legal disclosure obligations are jurisdiction- and framework-dependent and fall outside the scope of this concept as defined here.

Who it's relevant to

AI Governance and Compliance Officers
Those responsible for organizational transparency practices need to determine where AI disclosure is appropriate across customer-facing and internal systems, and to document the rationale for disclosure design choices. Because who, what, and when to disclose remain context-dependent, this group typically treats disclosure as a policy area requiring ongoing judgment rather than a fixed control, and should avoid assuming disclosure automatically strengthens trust.
Product and Customer Experience Teams
Teams building chatbots and other AI-mediated customer interactions face the practical task of notifying users that they are interacting with AI or that AI informs a decision affecting them. Evidence that disclosure can reduce trust in some settings makes framing, timing, and placement design decisions rather than mechanical additions, and these teams often own the resulting user-facing implementation.
Researchers, Academics, and Educators
In writing and research contexts, structured approaches such as the AID Framework and institutional AI disclosure statements provide methods for documenting how AI was used—including uses that might otherwise go unnoticed. This audience benefits from consistent, machine- and human-readable disclosure conventions to support academic integrity and reproducibility.
Legal and Policy Specialists
Professionals interpreting transparency obligations need to map disclosure practices to applicable requirements, which are jurisdiction- and framework-dependent and out of scope for this general definition. They should distinguish between disclosure as a voluntary governance practice and any specific legal duties, verifying the latter against the relevant instruments rather than assuming a universal standard.

Inside Disclosure of AI Interaction

Notice of AI Involvement
A statement informing a person that they are interacting with an AI system rather than a human, or that content or a decision was generated or materially shaped by an AI system. The form, timing, and prominence of such notice vary by framework and context.
Scope of the Interaction Disclosed
Clarification of what the AI is doing in the interaction, for example whether it is answering questions, generating content, or contributing to a decision. Disclosure practices differ on how much functional detail must accompany the basic notice.
Trigger Conditions
The circumstances under which disclosure is expected, which typically depend on context such as direct human-facing interaction, generation of synthetic media, or automated decision-making. Whether a given trigger is a legal obligation, guidance, or a voluntary practice depends on the applicable jurisdiction and framework.
Timing and Placement
When and where the disclosure is presented, such as at the outset of an interaction or alongside AI-generated output. Many practices favor disclosure that is available before or at the point a person could be misled about the nature of the counterpart.
Audience and Accessibility
Consideration of who receives the disclosure and whether it is presented in a manner they can reasonably understand. Practices commonly address plain language and accessibility, though specific requirements vary by context and are not uniform across frameworks.
Relationship to Broader Transparency Obligations
Disclosure of AI interaction is one component of wider transparency and governance measures and typically sits alongside, rather than replacing, obligations such as explanation of automated decisions or documentation of model behavior.

Common questions

Answers to the questions practitioners most commonly ask about Disclosure of AI Interaction.

Does disclosing that a user is interacting with an AI system make the system compliant with applicable transparency obligations?
Not necessarily. Disclosure of AI interaction is one measure that may address certain transparency expectations, but it is not equivalent to full compliance with any given framework. Depending on the jurisdiction and use case, applicable obligations may also touch on data handling, documentation, human oversight, and record-keeping. Disclosure typically addresses only the specific expectation that a person be made aware they are dealing with an AI system rather than a human, and should not be treated as a substitute for a broader compliance assessment. Where the precise scope of an obligation is uncertain, professionals should confirm it against the governing text rather than assume disclosure alone is sufficient.
Is disclosure of AI interaction a universal legal requirement for all AI systems?
No. Whether disclosure is required, and in what form, depends on the jurisdiction, the type of system, and the context of use. Some frameworks contemplate disclosure expectations for particular categories of interaction, while others treat it as a voluntary transparency practice or address it only in specific sectors. Treating disclosure as a blanket obligation across all systems and all jurisdictions would overstate its scope. As commonly framed, the specific triggers, exemptions, and manner of disclosure vary, so applicability should be determined for each system against the relevant governing instrument.
At what point in a user interaction is disclosure of AI involvement typically expected to be provided?
In many frameworks and internal policies, the aim is for disclosure to be provided at or before the point where a reasonable person could otherwise mistake the AI for a human, so that awareness precedes reliance. Practically, this often means presenting the disclosure at the outset of the interaction rather than only on request or after the exchange has concluded. The precise timing considered adequate can depend on the channel and use case, and specific requirements should be confirmed against the applicable framework rather than assumed to be a single fixed moment.
How can an organization document that AI interaction disclosures were actually presented?
Organizations commonly retain evidence such as records of the disclosure text used, version histories of that text, configuration settings that control when disclosures appear, and logs indicating that disclosures were rendered during interactions. Maintaining this documentation can support governance functions and reviews by second-line or third-line-of-defense teams. The appropriate retention period and level of detail typically depend on internal policy and any applicable record-keeping expectations, which should be confirmed for the relevant context rather than assumed.
Who within an organization is typically accountable for implementing and maintaining AI interaction disclosures?
Accountability is often distributed across lines of defense. Product, design, or operational teams (commonly associated with the first line) frequently implement and maintain the disclosure in the interface, while governance, compliance, or risk functions (often associated with the second line) may set policy and monitor adherence, and internal audit (the third line) may independently review effectiveness. This division of responsibility relates to AI governance structures rather than to model risk measurement itself, and the specific allocation varies by organizational design and should be defined in internal policy.
How can disclosure practices be kept consistent when AI is embedded across multiple channels or products?
Consistency is commonly supported by centralizing approved disclosure language, applying it through shared configuration or design components, and reviewing each channel where AI interaction occurs so that no channel is inadvertently omitted. Change-management and version-control practices can help ensure that updates propagate across channels and that older or inconsistent wording is retired. Because disclosure requirements and user expectations can differ by context and channel, organizations typically confirm that a single approach remains appropriate for each channel rather than assuming uniform wording fits every setting.

Common misconceptions

A single disclosure requirement applies universally to all AI interactions.
Disclosure expectations are not uniform. Whether disclosure is a binding legal obligation, guidance, or a voluntary practice depends on the jurisdiction, sector, and the nature of the interaction. Requirements that apply to consumer-facing chatbots or synthetic media in one framework should not be assumed to apply everywhere.
Disclosing that AI is involved satisfies all transparency and governance obligations.
Notice of AI involvement is a distinct and narrow component. It typically does not, on its own, address related concerns such as explaining how an automated decision was reached, documenting model risk, or providing recourse. These are separate measures that commonly coexist with disclosure.
Providing disclosure removes the risk that a person is misled or harmed by the AI interaction.
Disclosure is a measure that reduces the risk of a person being misled about the nature of their counterpart; it does not eliminate risk. It does not by itself guarantee the accuracy, fairness, or appropriateness of the AI's output, which remain subject to separate controls.

Best practices

Determine which specific obligations, guidance, or voluntary standards apply to your context and jurisdiction before assuming a particular disclosure form is required, and document that scoping analysis.
Present disclosure clearly and early enough that a reasonable person is not misled about whether they are interacting with an AI system, and use plain, accessible language for the intended audience.
Distinguish disclosure of AI interaction from adjacent transparency measures such as explanation of automated decisions, and treat them as complementary rather than interchangeable controls.
Tailor the level of functional detail to the interaction, describing what the AI is doing where that helps the recipient understand the nature of the interaction.
Maintain documentation of disclosure design decisions, trigger conditions, and placement so that governance, second-line, and audit functions can review them.
Periodically review disclosure practices as regulatory treatment evolves, and avoid presenting proposed or emerging requirements as settled obligations.