Chile's AI framework bill has cleared the Chamber of Deputies and is now moving through the Senate. The initiative is still under debate and hasn't come into force yet, but the map of requirements, oversight, and penalties, which can already be translated into pesos, will mark a before and after for the Chilean corporate environment.
Waiting for the regulation to be published in the Official Gazette before reacting is a management mistake. Below, we break down what this regulatory framework actually means, what happens if your company isn't aligned, and the key steps to get ahead of it in time.
Classifying AI by risk level
The bill doesn't regulate AI as a single block: it splits uses by the potential harm they can cause, across four levels modeled closely on the European approach.
- Unacceptable risk (banned without exception): subliminal manipulation, biometric categorization using sensitive data, generic social scoring, real-time remote identification in public spaces, emotion recognition.
- High risk (mandatory registration, auditing, and transparency): credit scoring, personnel selection, medical diagnosis. If the system fails to comply, the operator must deactivate it, pull it from the market, or suspend it immediately.
- Limited risk (transparency obligation): uses with no significant risk of manipulation or error. Users must be able to know they're interacting with an AI system.
- No evident risk (free use): most productivity and internal automation tools fall here, with no additional obligations.
If your company uses AI to evaluate candidates, approve loans, or support medical diagnoses, it falls straight into the high-risk category, the most demanding one in the bill. It's worth looking through that lens at any tool already in production, including one that arrives bundled inside a third-party vendor's software.
The fines
The bill sets three tiers of administrative sanctions. Translated into pesos, using the August 2026 UTM value ($71,649):
- Minor infractions: up to 5,000 UTM (approx. $358 million pesos).
- Serious infractions: up to 10,000 UTM (approx. $716 million pesos).
- Very serious infractions: up to 20,000 UTM (approx. $1,433 million pesos).
The exact amount within each tier isn't fixed. The authority must weigh the organization's size and annual sales, the economic benefit obtained from the infraction, repeat offenses, the number of people affected, and the degree of cooperation with the investigation, among other factors. In practice, this means a small business and a large company don't risk the same amount, but both stay on the radar.
Who enforces it?
The bill creates the National Artificial Intelligence Commission, responsible for authorizing high-risk systems, keeping their registry, and ruling on serious incidents. The Ministry of Science, Technology, Knowledge, and Innovation acts as the general coordinator for promoting ethical AI development in the country.
One nuance worth following: the bill includes differentiated standards based on company size, a sign that Congress is aware not every Chilean company has the same compliance muscle to meet these requirements from day one.
Chile looks at Europe, but doesn't copy it
Both frameworks share the logic of classifying systems by risk level and ban similar practices: subliminal manipulation, exploitation of vulnerabilities, social scoring. The difference is in the detail: the European AI Act is considerably more specific in its list of high-risk systems, while the Chilean text leaves more room for interpretation and for a future regulation that doesn't exist yet.
While the European Union is closing out a horizontal framework built over years, Chile is still negotiating its own in committee: the text can still change before it becomes law.
What to do while it's being debated
The law isn't in force yet, but the map of obligations is already clear enough to get ahead of it:
- Inventory the AI systems your company already uses, both in-house and from vendors.
- Classify them against the draft risk categories: does any of them touch credit, personnel selection, or health?
- Review what documentation you have on the origin and workings of each model in use.
- Follow the bill's progress in the Senate. The Future Challenges Committee is where the details are being defined.
The advantage of moving before the law takes effect isn't just avoiding a fine that doesn't exist yet. It's arriving at the moment of enforcement with the work already done, instead of reacting.
