Data Minimization

Data collected for one explicit purpose cannot be repurposed for incompatible secondary uses, such as training a commercial AI model, without obtaining new consent. This principle is a legal constraint that directly limits the scope of data ingestion.

• Legal Basis: Enshrined in GDPR Article 5(1)(b).
• Mechanism: Requires a direct link between the stated purpose at collection and the processing activity.
• AI Impact: Prevents 'scope creep' where customer support logs are later used for unsupervised model pre-training.

Personal data must be kept in an identifiable form for no longer than is necessary for the purposes for which it is processed. This mandates the systematic deletion of stale data from AI training corpora once the original purpose expires.

• Retention Policy: Organizations must define standardized timelines for archiving and purging data.
• AI Hygiene: Ensures opted-out or obsolete data is removed from vector databases and feature stores.
• Compliance: Directly supports the Right to Erasure by preventing indefinite data hoarding.

The process of creating a comprehensive, visual record of all data assets flowing through an organization. This is a prerequisite for minimization, as you cannot protect what you cannot see.

• High-Risk Identification: Pinpoints datasets inadvertently exposed to AI crawlers or unauthorized training pipelines.
• RoPA Alignment: Feeds directly into the mandatory Record of Processing Activities required under GDPR.
• Automated Discovery: Modern tools scan structured and unstructured data stores to detect personal information.

A legal provision under GDPR Article 21 granting individuals the absolute right to object to the processing of their personal data based on legitimate interests, including profiling. This can be invoked directly against AI training.

• Absolute Right: For direct marketing, the objection must be honored without exception.
• Burden of Proof: The data controller must demonstrate 'compelling legitimate grounds' to override the objection.
• AI Context: Users can object to their public data being scraped for generative model fine-tuning.

A machine-readable legal instrument, often leveraging Creative Commons-style frameworks, that explicitly grants or denies specific usage rights for data, including permissions for AI training and computational analysis.

• Granular Permissions: Specifies if data can be used for Text and Data Mining (TDM).
• Automated Compliance: Allows crawlers to parse usage rights without human intervention.
• Standardization: Moves beyond vague 'all rights reserved' statements to explicit, actionable licenses.

A standardized, auditable digital record provided to a data subject that details the specifics of a consent transaction. It serves as proof that permission was or was not granted for AI model training.

• Audit Trail: Links a specific individual to a specific consent version and timestamp.
• Key Fields: Records the purpose (e.g., 'AI Training'), the data categories, and the withdrawal mechanism.
• Interoperability: Designed to be portable across different Consent Management Platforms (CMPs).

A binding legal constraint that prohibits data collected for one explicit purpose from being repurposed for incompatible secondary uses, such as training a commercial AI model, without obtaining new consent. This principle directly enforces data minimization by preventing function creep.

• Requires a defined, explicit, and legitimate purpose before collection
• Prohibits repurposing customer support transcripts for LLM fine-tuning
• Enforced under GDPR Article 5(1)(b)

A data governance principle requiring that personal data be kept in an identifiable form for no longer than necessary for the specified purpose. This mandates the systematic deletion of stale data from AI training corpora once the original processing purpose expires.

• Forces periodic review of data retention schedules
• Prevents infinite accumulation in data lakes used for model training
• Complements automated data lifecycle management tools

Also known as the 'right to be forgotten', this legal right compels data controllers to delete personal data without undue delay. It poses a significant technical challenge for foundation models, requiring model unlearning techniques to remove the influence of specific data points from trained weights.

• Applies when data is no longer necessary for the original purpose
• Requires propagation of deletion requests to downstream processors
• Challenges the immutable nature of static training datasets

The process of creating a comprehensive, visual record of all data assets flowing through an organization. This is a prerequisite for data minimization as it identifies high-risk datasets that may be inadvertently exposed to AI crawlers or unauthorized training pipelines.

• Catalogs data sources, flows, and storage locations
• Identifies redundant, obsolete, or trivial (ROT) data for deletion
• Essential input for a Record of Processing Activities (RoPA)

A technical specification, often implemented via the robots.txt parser or HTTP headers, that allows rights holders to communicate a reservation of rights for text and data mining. It operationalizes data minimization by instructing compliant crawlers not to ingest content for AI training.

• Extends the Robots Exclusion Protocol for AI-specific directives
• Enables granular path exclusion for sensitive directories
• Relies on voluntary compliance by crawler operators

A cryptographic technique that injects calibrated statistical noise into datasets or model training processes to prevent the extraction of individual records. It provides a mathematical guarantee of data minimization by ensuring that the presence or absence of a single data point does not significantly alter an output.

• Quantified by the privacy loss parameter epsilon (ε)
• Allows training on sensitive data without exposing individual contributions
• Used by the US Census Bureau and major technology platforms