Methodology

Methodology

How VIGÍA measures AI regulatory action in Mexico.


Scope of the bilingual content

The INTERFACE is fully bilingual (navigation, headings, labels, severity map, taxonomy names, and this methodology). The substantive NARRATIVE CONTENT —the DOF analysis— stays in Spanish under both languages, because translating it would require document processing that is out of scope. Where Spanish content appears under the English language, a discreet notice is shown. This is a deliberate, documented decision, not a defect.

Source content shown in its original language (Spanish).


Functional, not nominal, detection

A finding is detected by its function, not because it explicitly mentions AI. That is why we distinguish explicit from implicit findings: much of the relevant regulation (silent AI) never names artificial intelligence yet regulates it in practice.

545 Implicit - Technology
120 Explicit
95.6% Regulatory gap rate

Verbatim citation principle

Every finding keeps the original fragment of the document. The citation is language-invariant (always in Spanish) so that anchoring against the source text stays valid in both interface languages.


6-level severity scale

Each finding is ordered on a six-level scale, including the intermediate Medium-High level. Severity is stored as a canonical ordinal to allow consistent sorting and aggregation.

# Severity
0 Low
1 Medium
2 Medium-High
3 High
4 Very High
5 Extreme

Risk taxonomy (R1–R15)

VIGÍA classifies AI risks into 15 categories. The taxonomy is transversal: it applies to every DOF finding, which lets us aggregate and compare them on a common basis.

R Category DOF findings
R1 Malfunctions & Errors 233
R2 Discrimination & Bias 301
R3 Privacy Invasions 406
R4 Disinformation & Deepfakes 44
R5 Copyright Infringement 27
R6 Worker Exploitation 11
R7 Labor Displacement 72
R8 Reduced Social Connection 34
R9 Authoritarian Surveillance 177
R10 Concentration of Power 74
R11 Bioterrorism Facilitation 6
R12 War Escalation 12
R13 Gradual Loss of Control 24
R14 Agent Misalignment 0
R15 Unknown & Emerging Risks 45

View all


Source corpus

Analysis of the Official Gazette of the Federation: 153 documents published between November 2025 and June 2026, with 665 findings. The original corpus is not modified; ingestion is read-only.

153 DOF documents
665 findings
1,466 risk tags

Sources and reproducibility

The corpus totals are verifiable and reproducible from the source files. The database keeps the structural hooks for future analysis without requiring schema migrations.