Algorithmic law: who is liable when a machine decides

An algorithm rejects your loan. Another filters your CV before a human reads it. A third decides which posts you see and which you do not. The question the…

2026-05-29 · AI and data compliance

An algorithm rejects your loan. Another filters your CV before a human reads it. A third decides which posts you see and which you do not. The question the law can no longer ignore is: who is liable when that system gets it wrong, discriminates or affects you unfairly? Chile's AI bill and Ley 21.719 are starting to answer. Here is what you need to know.

When the decision-maker is a system

Automating decisions is not new, but its current reach is. Today it is common for artificial intelligence systems to take part in decisions on credit, hiring, allocation of public services, surveillance and risk assessment. The efficiency they offer is real, but an equally real gap opens up: if the system gets it wrong or reproduces discrimination, who is responsible?

Traditional law assigns liability to people and organisations. When the decision-maker is a model trained on millions of data points, that chain of responsibility becomes opaque. That is exactly what so-called "algorithmic law" is trying to resolve.

The concrete problems the regulation seeks to address

Bias and discrimination: a model trained on historical data can reproduce discriminatory patterns of gender, ethnicity or socio-economic class, even though nobody explicitly programmed it to. If the historical hiring dataset favoured men, the algorithm will tend to do the same.
Opacity (the "black box"): many modern AI systems cannot explain in human terms why they took a decision. That is a problem when the decision affects you directly and you want to challenge it.
Diffuse liability: is it the person who trained the model, the one who deployed it or the one who used it to take the decision? Without clear rules, the victim has nobody to turn to.
The right to human review: in decisions affecting important rights, the regulatory trend is that the algorithm's output is not enough: there must be a human able to review and correct it.

Chile's AI bill: how it classifies risks

In October 2025 Chile's Chamber of Deputies approved a bill on Artificial Intelligence that is currently still going through the Senate. Its approach is similar to the European AI Act: classifying AI systems by the risk they pose to fundamental rights.

Unacceptable risk: uses directly prohibited as incompatible with fundamental rights (for example, mass social scoring systems or indiscriminate biometric surveillance).
High risk: systems affecting sensitive decisions —health, credit, employment, education, justice— which require greater transparency, documentation and human oversight.
Limited risk: systems with minor potential for manipulation, subject to basic transparency obligations (for example, telling the user they are talking to a bot).
No obvious risk: low-impact uses with minimal restrictions.

The bill also creates a Technical Advisory Council on AI under the Ministry of Science, and sets fines of up to 20.000 UTM for those breaching the rules. You can follow its progress at the Chamber of Deputies.

How it connects with Ley 21.719 on personal data

Algorithmic decisions are almost always based on personal data. That is why Ley 21.719, which comes fully into force on 1 December 2026, is complementary: it provides that people have the right not to be subject to decisions based solely on automated processing where those decisions produce significant legal effects or similarly affect them. In those cases, the data subject has the right to obtain human intervention, to express their point of view and to challenge the decision.

In practice, this means that if a bank rejects your credit application using an automated model, you have the right to ask for a human to review that decision.

A concrete example: the automatic CV filter

Many companies use ATS (Applicant Tracking Systems) and, increasingly, AI models that score CVs before they reach a person. If the model was trained on data from previous employees —mostly men, from a certain university, with a certain profile— it will tend to filter out candidates who do not fit that mould, even if they are equally capable.

Under the framework being built in Chile, such a system would be "high risk" (an employment decision with significant impact). That would entail compulsory documentation, periodic audits and the possibility for the rejected candidate to request an explanation and human review.

What this means in practice

If you use automated systems to take decisions about clients or workers: the regulatory direction requires you to be able to explain, oversee and document them. Getting ahead today is cheaper than adapting under pressure.
As an affected person: you have the right to know a decision was automated, to ask for an explanation and, in significant cases, to demand human review. Those rights are already partly in Ley 21.719 and will be reinforced by the AI law.
For legal and technology professionals: the intersection of AI regulation, data protection and civil liability is one of the fastest-growing fields in legal practice over the coming years.

In summary

Algorithmic law responds to a real need: that decisions affecting us —even when taken by a machine— be explainable, supervised and open to challenge. Chile is moving in that direction with the AI bill and Ley 21.719. For companies, the message is clear: transparency and human oversight are no longer optional. For individuals, knowing these rights means no longer being at the mercy of systems nobody explains to them.

Sources: AI bill — Chile's Chamber of Deputies; Ley 21.719 — BCN. Informational content: it does not replace professional advice.

— Mindset & Code · Legal-Tech Chile · law and technology, explained simply.