Here is a question most people have never thought to ask: when an algorithm decides something important about you, do you have a right to know? Not a right to a good outcome — just a right to know that a machine was the one deciding, and why.
The answer is quietly becoming ‘yes’. Across the United States and Europe, an old legal idea is being applied to a new technology: the idea that you are entitled to information when a decision affects your life, and that a system you cannot see should not be allowed to judge you in the dark.
The law already had the words
The striking thing is how little new law was actually needed. The United States has had anti-discrimination statutes for decades — protections against discrimination in employment, credit, housing and education. The courts have now made clear, with remarkable speed, that those laws apply to algorithms. A machine making a hiring decision is subject to the same rules as a human making it. Judges have explicitly refused to create a ‘software exception’: the fact that the decision was made by software does not reduce the employer’s liability for discriminatory outcomes.
That refusal matters enormously. It means the entire edifice of civil rights law — built over fifty years for human decision-makers — now also governs machines. An algorithm that rejects loan applications on a biased basis can violate the same fair-lending law as a biased loan officer. There is no new statute required, and there is no escaping to the excuse that ‘the computer did it’.
Think about what that means for how companies build software. If the algorithm is as accountable as the human, then the process of building it — the data it was trained on, the thresholds it uses, the way its decisions are tested — becomes a matter of legal exposure. Bias is no longer a technical defect to be patched; it is a liability to be governed. That single reframing changes the incentives of the entire industry.
The new obligations, in practice
What does this actually require of organisations? The pattern across the new rules is remarkably consistent. New York City’s bias audit law requires companies using automated employment tools to conduct independent annual audits for bias. California’s automated decision-making rules demand pre-use notice, opt-out rights and access rights for significant decisions in hiring, lending, housing, education and healthcare. Colorado’s AI Act requires reasonable care to avoid algorithmic discrimination. Roughly a hundred state AI laws have been enacted, and the throughline of most of them is the same word: disclosure. Tell people when AI is talking to them, deciding about them, or generating what they see.
There is also a notably sharp edge growing in the case law. A federal court certified a class action treating an AI vendor as an agent of the employer — meaning the company that made the software can share legal responsibility for its discriminatory decisions alongside the company that used it. That is a new kind of liability, and it sends a message to the entire industry: building the tool does not exempt you from the consequences of what it does.
The practical consequence is that compliance is becoming a professional discipline. Organisations now need people who can explain their systems, document their testing, and answer for the boundary between what the machine decided and what a human oversaw. The titles are new, but the job is ancient: someone has to be responsible.
Why the right to know is the right foundation
It is worth thinking about why ‘notice’ is the organising principle rather than, say, a ban on automation. There are good reasons. A ban is blunt and would throw out enormous benefit. Notice is precise: it does not stop the machine from deciding; it stops the machine from deciding invisibly. The moment an organisation must disclose that a decision is automated, it must also be able to explain the decision. And the moment it must explain, it must understand its own system. Disclosure forces competence.
That is the deep argument for transparency as the legal foundation. A right to know is not just consumer protection; it is the pressure that keeps the entire ecosystem honest. Organisations that must answer for their algorithms will test them, document them and supervise them in ways they never would if the algorithm could decide in silence. The obligation to explain is, in effect, an obligation to understand.
There is also a dignitarian argument hiding in the principle, and it is worth naming. Being judged by a system you cannot see or question is a form of powerlessness that touches something fundamental. The right to know is, at bottom, a claim that you are owed a reason — that your fate, however it is decided, is not permitted to be a black box to you. That is the same claim that underpins due process, and it is being extended, decision by decision, to the machines.
The limits nobody should hide
Be equally honest about the limits. A right to know is not a right to a good outcome. Disclosing that a machine decided a loan application does not guarantee the loan. The explanation you receive may be genuine or it may be a form letter reciting the factors the algorithm considers — useful but not equal to the algorithm’s actual reasoning. Enforcement is uneven, resources are thin, and in many jurisdictions the right exists on paper more than in practice.
And there is a genuine tension the law has not resolved: the tension between your right to understand a decision and the organisation’s claim that its algorithm is a trade secret. In the coming years, courts will be asked again and again how much of a machine’s reasoning an individual is entitled to see. The answer will shape whether the right to know is real, or just well-advertised.
The direction, though, is settled. The idea that you should not be judged by a machine you cannot question has crossed from a principle into statutes, regulations and court rulings. It is now a right with teeth — imperfectly enforced, unevenly applied, but real. The next decade will be spent deciding what ‘explainable’ actually means in practice, case by case, statute by statute. The era in which an algorithm could decide your fate with no explanation, no recourse and no obligation to justify itself is drawing to a close. You may not always win the decision. But you are, slowly and unmistakably, winning the right to know why.