The Right to Know You Are Talking to a Machine

A growing share of the messages, decisions and images that shape daily life are produced by systems that do not announce themselves. A chatbot answers a complaint; an algorithm sorts a loan application; a synthetic video circulates as though it were footage. A distinct family of AI laws responds to this quiet substitution with a simple demand: that people be told. This article examines the access to information and transparency mandates approach, and why disclosure is the foundation on which every other protection rests.

Introduction

Algorithmic transparency is among the most common principles in the world's AI ethics frameworks and bills (UNESCO, 2026). The reason is structural rather than merely moral. A person cannot question, contest or trust a decision they do not know was made by a machine. Transparency mandates convert that abstract principle into concrete obligations to disclose - about how a system was built, what it does, and when it is operating.

From Principle to Obligation

Early transparency duties targeted the state. France's Law No. 2016-1321, known as the law for a Digital Republic, requires public bodies to publish the rules defining the main algorithmic processes used when those processes form the basis of individual decisions (UNESCO, 2026). The obligation treats a citizen's right to understand public decision-making as continuous with the older right of access to information.

More recent laws extend disclosure to the private sector and to content itself. Article 50 of the European Union's AI Act requires that systems intended to interact directly with people be designed so that users are informed they are dealing with an AI, and that AI outputs be marked in a machine-readable format as artificially generated. It obliges anyone who produces a deepfake to disclose that the content has been artificially generated or manipulated, and imposes a comparable duty on AI-generated text published to inform the public on matters of public interest (UNESCO, 2026). South Korea's AI Act, through Article 31, adds that operators of high-impact or generative AI must notify users in advance, label generative outputs, and clearly flag synthetic voices, images or videos that are difficult to distinguish from reality (UNESCO, 2026).

The First Rung, Not the Last

Transparency is valuable less for what it achieves alone than for what it makes possible. The report frames it as an enabler - a necessary first step toward the further objectives of explainability and accountability (UNESCO, 2026). The relationship is sequential, as the figure above suggests: a person must first know an AI was involved before they can ask why it produced a particular output, and they must be able to understand that output before anyone can be held responsible for it. Remove the bottom rung and the ladder does not stand.

The Psychology of Calibrated Trust

Disclosure also does something subtler: it lets trust find its correct level. Research by John Lee and Katrina See on trust in automation established that the goal is not maximal trust but appropriate reliance - trust calibrated to a system's actual capabilities, so that people neither over-rely on a flawed tool nor abandon a sound one (Lee & See, 2004). Calibration is impossible in the dark. A user who does not know a system is automated cannot adjust their confidence to its competence, and an undisclosed machine invites precisely the miscalibration - misplaced trust, or misplaced suspicion - that transparency is meant to prevent.

There is an ethical dimension the psychology only sharpens. When artificiality is concealed, a person's trust is not earned but extracted, granted on a false premise. The right to know one is speaking to a machine is therefore not a technicality. It is the precondition for every judgement a person might reasonably want to make about the machine - and, as the next article argues, for holding anyone to account when the machine causes harm.