RIPD approves AI data standards without adding new general obligations
The standards treat public availability as insufficient for AI training and reject human review that merely rubber-stamps an automated decision.
The Ibero-American Data Protection Network (RIPD) unanimously approved its Ibero-American Standards on the use of personal data in artificial intelligence systems on 12 November 2026. The Spanish Data Protection Agency (AEPD), which holds the RIPD's permanent secretariat, said the international standards explain how existing data protection rules apply to AI used by companies and public administrations. They do not create a parallel legal regime or new general obligations.
That distinction matters for compliance teams. The standards say AI inferences, predictions and profiles may constitute personal data where they relate to identified or identifiable people, and require organisations to account for the uncertainty of those outputs and ensure that human review is genuine. The standards also cover generative and agentic AI, biometrics, emotion recognition and neurodata.
What the standards clarify
The standards address data that an AI system infers rather than data collected directly from an individual. Organisations should account for the uncertainty of an inference and should not present a prediction as more certain than the evidence supports. If a system infers health, ideology or sexual orientation from apparently neutral information, the enhanced safeguards for sensitive data apply.
Public availability is not, by itself, permission to use personal data to train an AI system. The standards require an appropriate legal basis and consideration of what people could reasonably have expected when they published the information.
Transparency requires more than telling people that AI is being used. Organisations should explain in understandable language:
- why the system is used;
- what role it plays in the relevant process; and
- what consequences it may have for the individual.
Trade-secret protections do not remove that right to information.
The standards also address models that have already been trained. People should be able to find out whether their data was used in an AI system and, where an automated decision affects them, challenge the conclusion, exercise their applicable rights and demand real human intervention. A reviewer who simply endorses the system's recommendation does not provide effective human intervention for these purposes.
Where a model cannot fully delete personal data, the organisation should explain and justify that limitation, inform the individual and take measures such as excluding the data from future training. The standards describe systems designed to deceive people, exploit their vulnerabilities or apply discriminatory social scoring as incompatible with data protection.
Providers and users share defined responsibilities
A provider using customer data to train or improve its own models needs an appropriate legal basis and must meet the corresponding data protection requirements. That issue is separate from the provider's role in supplying a model to a customer.
An organisation using a third-party AI system remains responsible for processing where it determines the essential purposes and means, even if it neither developed nor trained the model. The standards therefore leave deployment responsibility with the organisation controlling the relevant processing rather than allowing procurement of an external model to transfer it automatically.
They also emphasise data protection by design, governance, data quality, human oversight and risk assessment and management. Measures should be adapted to the characteristics and risks of each processing activity, rather than applied as an identical control set to every AI use case.
In Atlas's view, the standards are best read as an implementation framework for existing data protection duties, not as a new standalone compliance layer. For firms operating across Ibero-American markets, the practical work is to map existing AI inventories, impact assessments, notices, data-source controls, model governance and human-review procedures against these interpretations.
The standards were developed through drafting, consultation and technical deliberation by RIPD member authorities, following an initial version prepared by consultant Marcos Sánchez Acevedo. The European Union financed the work with the Foundation for the Internationalization of Public Administrations (FIAP) under the EU-LAC Digital Alliance.
They complement the RIPD's existing Standards for the Protection of Personal Data for Ibero-American States, which already addressed AI, automated decisions, bias prevention and algorithmic governance.
Spain's enforcement context
The approval is a policy and interpretive instrument, not an AEPD enforcement decision.
The standards' significance is as a common reference for authorities, legislators, companies and administrations applying existing data protection rules to AI across the RIPD's more than 600 million people.
Permanent link to this Atlas analysis
Related Atlas analysis
Continue with coverage connected by market, topic and operating context.
- SpainSpain approves ONCE’s €3 Eurojackpot changes from 29 MayThe first new-rules draw is 1 June 2027, and every Spanish combination must include the €0.50 Number of Luck.
- TurkeyTürkiye central bank raises loan limits and remunerates long-term-liability reservesEight-week caps rise by 0.5 percentage points across consumer, SME and non-SME lending categories.
- IndiaTRAI opens consultation on responsible AI in telecom until 27 NovemberThe draft applies India’s seven-principle AI governance framework to a sector linked to critical information infrastructure, national security and emergency response.
Continue with Atlas
Move from this development into the relevant research, comparison and workflow.
- Gambling compliance software guideSee the research, monitoring, ownership and evidence a useful platform should connect.
- Atlas vs VixioCompare gambling research, change monitoring, workflow, technical work and evidence.
- Monitoring and alertsFollow a relevant change from detection through accountable implementation.