Data protection update
Ibero-American Standards on the Use of Personal Data in Artificial Intelligence Systems
Colombia's data protection authority (SIC) announced the Ibero-American Standards on personal data in AI systems. They are not binding, but they will guide how Law 1581 of 2012 is applied: inferences can be personal data, publicly accessible data is not “public data”, and hiring an external AI tool does not shift responsibility.
On November 17, 2026, the Superintendence of Industry and Commerce (SIC), as Colombia's personal data protection authority and a member of the Ibero-American Data Protection Network (RIPD), announced the publication of the Ibero-American Standards on the Use of Personal Data in Artificial Intelligence Systems (the “Standards”). They were unanimously approved by the Network's authorities and prepared in collaboration with the Foundation for the Internationalization of Public Administrations (FIAP).
The Standards are not, in themselves, a binding rule; however, the SIC presents them as a reference instrument for authorities, legislators, controllers and processors, AI system providers and other public and private organizations. For this reason, they are likely to guide how the SIC interprets and enforces Law 1581 of 2012 when AI tools are used.
What you should know
Scope of the Standards
- They apply to the processing of personal data and its purposes, not to AI systems or models or to the technology that supports them.
- Each phase of the processing must be analyzed: development, training, evaluation and improvement of the system.
- They also cover the personal data that the systems generate (inferences, profiles, predictions), even if AI intervenes in only part of the processing.
Main clarifications announced by the SIC
- The inferences and profiles generated by an AI system can be personal data. If they reveal sensitive aspects (for example, health or political orientation), the enhanced protection for sensitive data applies, which in Colombia requires, among other things, explicit authorization and that answering be optional.
- The fact that data is publicly accessible is not, by itself, enough to use it to train AI systems. It is important not to confuse “accessible on the internet” with “public data” within the meaning of Law 1581 of 2012 and Decree 1377 of 2013: the fact that data can be consulted online does not remove the need for a legal basis that legitimizes its use.
- Stating that AI is used does not necessarily satisfy the duty to inform. It must be explained what it is used for, what role it plays in the processing and what consequences it may have for the data subject.
- Individuals can challenge automated decisions and demand genuine human intervention. There is no genuine human intervention if the reviewer merely endorses what the AI proposes.
- Whoever uses third-party AI systems remains responsible for the processing whose essential purposes and means it determines. Hiring an external tool does not shift responsibility to the provider.
- AI systems designed to manipulate people, exploit their vulnerability or score them socially in a discriminatory way are incompatible with the Standards.
Where to find the Standards
The Standards can be consulted on the RIPD website (www.redipd.org), which publishes them in Spanish, English and Portuguese, and on the website of the Spanish Data Protection Agency (www.aepd.es), which acts as the Network's permanent secretariat. It is a guidance document that complements the Data Protection Standards for the Ibero-American States without lowering their level of protection, and it is organized into a preamble and nine (9) chapters.
Recommendations
- Inventory the AI tools (in-house and third-party) and identify which personal data they process, in which phase and for what purpose.
- Verify that each processing activity and phase has a legitimate basis.
- Review contracts with AI providers to define roles and to expressly regulate whether the provider may use the organization's data to train its models.
- Update the Personal Data Processing Policy and privacy notices to explain to data subjects what AI is used for, what role it plays and what consequences it may have; merely stating that it is used is not enough.
- Establish genuine human review of automated decisions with relevant effects, and a channel for data subjects to challenge them.
- Assess the impact before deploying AI systems.
- Incorporate AI-specific threats into security and incident protocols.
We are here to help
We hope this information is clear. Please do not hesitate to contact us if you have any questions or concerns, or if you would like to schedule a meeting to discuss this topic in more depth.
Do you need to review a specific case?