ChileBoletín 16.821-19, clasificación de riesgo y prohibición de sistemas de IA de riesgo inaceptable
An AI prohibited practices rule binding public and private bodies.
- How the hook was established
- express
- What makes it apply
- Market targeting, Operator establishment, Place of effect
- Audit expectation
- none
What it requires
- Before distributing, placing on the market or putting into service an AI system in Chile, determine whether its intended use is an unacceptable-risk use listed in Article 7 (subliminal manipulation, exploitation of personal characteristics, categorization based on sensitive personal data, generic social scoring, real-time remote biometric identification in publicly accessible spaces, untargeted extraction of facial images, or emotion inference in criminal law enforcement, border management, workplaces and educational centers), since such systems are prohibited, subject to the therapeutic exceptions in Article 7(1) and 7(3).
These lines are LexLint's reading of the law's text.
Who enforces it
Settledness
Article numbers follow the full text of the bill as approved by the Chamber of Deputies' Comisión de Futuro, Ciencias, Tecnología, Conocimiento e Innovación in its second report.
- As of
- Open questions
- Does the prohibition in Article 7(5) of "sistemas de IA utilizados para el análisis de imágenes de vídeo en espacios de acceso público que emplean sistemas de identificación biométrica remota en tiempo real" ("AI systems used to analyze video images in publicly accessible spaces that employ real-time remote biometric identification systems") reach use by public authorities for criminal investigation, given that Article 7(5) states no exception and Article 2 excludes only national defense, pre-market research and testing, and open-source components?
- Does Article 7(7), which bans systems that "pretenden inferir las emociones de una persona natural" ("claim to infer the emotions of a natural person") in workplaces and educational centers, reach a system that infers emotion from text alone, given that Article 3(15) describes emotion recognition as detecting emotions "a partir de sus datos biométricos y otras técnicas" ("from their biometric data and other techniques")?
- Does a provider that develops or tests a system within a use banned by Article 7 breach the prohibition during development, given that Article 6 counts "desarrollo, prueba y validación" ("development, testing and validation") as "uso" ("use") while Article 2(b) excludes "actividades de investigación, pruebas y desarrollo sobre sistemas de IA" ("research, testing and development activities on AI systems") carried out before placing on the market?
What it makes you log
Logging duty
Articles 6 and 7 classify AI uses by risk and prohibit the unacceptable-risk uses listed in Article 7; they state no duty to create, keep or produce a log or record, and the record duty sits in Article 9(4) for high-risk systems.
- Kind
- None
- As of
What this law does
Article 6 of this bill classifies the uses of AI systems by risk into four categories (uso de riesgo inaceptable, uso de alto riesgo, uso de riesgo limitado, usos sin riesgo evidente). Article 6 also declares prohibited the distribution, placing on the market or putting into service of AI systems intended for unacceptable-risk uses, defined as uses incompatible with respect for and guarantee of people's fundamental rights.
Article 7 lists those uses: subliminal manipulation, exploitation of personal characteristics to generate harmful behavior, categorization of persons based on sensitive personal data, generic social scoring, real-time remote biometric identification in publicly accessible spaces, untargeted extraction of facial images, and inference of emotions in criminal law enforcement, border management, workplaces and educational centers.
The prohibition on subliminal manipulation does not apply to AI systems intended for therapeutic purposes carried out in accordance with law, which require the person's informed, specific and express consent. The prohibition on categorization based on sensitive personal data does not apply to AI systems intended for therapeutic purposes authorized on the basis of informed, specific and express consent, together with the relevant health authorization where applicable. The bill reached the Senate from the Chamber of Deputies for its second constitutional stage in October 2025.
When LexLint raises it
When your app profile says your app makes high-risk automated decisions.
Back to the example · Lint your app
Page updated .