Colombia - AI and Labor Rights (130/2023)

Overview

Proyecto 130/23 (Senate) — "Por medio de la cual se crea la armonización de la inteligencia artificial con el derecho al trabajo de las personas" — was presented to the Colombian Senate on 6 September 2023. The text proposed a focused regulatory framework governing the use of AI systems in employment contexts (selection, performance evaluation, discipline, task allocation, surveillance, remuneration and termination) to preserve workers’ rights, workplace health and non‑discrimination. The bill combined obligations on employers (public and private) with roles for sectoral authorities. Official project metadata and the radicated text are available from the Senate project repository: Senate project page (Project 130/23). Secondary analyses and legislative trackers summarise the bill’s structure and status (see CELE observatory and Foro GPP).

Definitions

The bill defines key concepts for its application framework: "artificial intelligence" (broadly as systems that perform tasks replicating cognitive functions), "AI system" (software or combination of software and hardware deployed to make or support decisions), "employer" (public or private entity with employment relationships or contractual relations), "affected individual" (worker, candidate, contractor), and risk level classifications for AI systems (e.g., null/low, limited, high, unacceptable). The draft text sought to align definitions with existing Colombian ICT and data protection law, while introducing employment‑specific definitions to delimit coverage and obligations.

Governance and Institutional Framework

Proyecto 130/23 assigns institutional responsibilities across multiple agencies. It tasks the Ministry of Labour with labor‑policy oversight and guidance on worker protections; the Ministry of Information and Communications Technologies (MinTIC) with technical standards and classification criteria; and the Superintendence of Industry and Commerce (SIC) with market surveillance, enforcement and a registry function for higher‑risk models. The bill also envisaged inter‑ministerial coordination and delegated powers to issue technical regulations and implementation protocols. For the official project metadata and listing of sponsors and commissions see the Senate project entry: Senate project page (Project 130/23). The design adopts a co‑regulatory approach — binding obligations in the law with delegated rulemaking for technical detail.

Key Focus Areas

The bill centers on several interlocking areas: (1) Transparency and disclosure — mandatory notices when AI is used in hiring, evaluation or termination, and mechanisms for affected workers to obtain meaningful information about algorithmic decision‑making; (2) Human oversight — a requirement that decisions materially affecting employment must be subject to human review and that human judgement prevails; (3) Risk classification — systems are categorized by expected risk to fundamental labor rights and stronger obligations attach to higher categories; (4) Impact assessments — employers must complete algorithmic impact assessments (AIA) and workplace health impact studies before deploying AI systems; (5) Employment protections — transitional measures, reskilling obligations, reassignment and minimum notice or compensation schemes where AI deployment causes displacement; (6) Data protection — explicit requirements on consent, minimization, security, and use limits for worker data in AI training; (7) Documentation and registries — technical dossiers, model cards and (for higher risk systems) registration with a supervisory authority; and (8) Sanctions and redress mechanisms — administrative penalties, individual remedies and rights of challenge. Analyses and critical commentaries on the bill and related proposals are available from public legal observatories: CELE and Foro GPP.

Implementation Framework

Implementation envisioned a staged approach: immediate application of transparency and human‑oversight duties upon enactment; phased obligations for registration, conformity assessment and third‑party audits tied to risk tiers; and delegated rulemaking for technical standards (data quality, security, evaluation metrics) to MinTIC and technical panels. Employers would be required to create internal governance structures (AI compliance officers, audit trails, incident reporting) and to engage in worker consultation processes and social dialogue when AI materially affects jobs. The bill anticipated guidance, templates and capacity building to be issued by responsible ministries and supervisory agencies to support compliance in small and medium employers.

Monitoring and Evaluation

Monitoring provisions included supervisory inspections, mandatory periodic reporting by covered employers on AI deployments that affect labor decisions, and a requirement that supervising agencies publish aggregated enforcement statistics and anonymized case studies. The SIC was assigned a central market surveillance role, and the Ministry of Labour a role in assessing employment impacts and in collecting information on training/reskilling outcomes. The bill also encouraged pilot programs and sandboxes for controlled testing under regulator supervision to build evidence and calibrate obligations.

Penalties, Liability, and Appeals

The proposed sanctions regime combined administrative fines and corrective orders for non‑compliance with disclosure, impact assessment, documentation and reskilling obligations. It contemplated civil liability channels for workers harmed by algorithmic decisions (remedies for discrimination, wrongful termination or health harms) and administrative appeal routes against regulator decisions. The bill proposed objective liability for certain high‑risk failures and left open the development of detailed fine schedules and procedural guarantees in delegated regulations.

Relationship to Other Instruments

Proyecto 130/23 was drafted in the same legislative cycle as other Colombian AI bills and intersects with existing legal frameworks: Law 1581 of 2012 (personal data protection), Ley 1341 of 2009 (TICs), the Labor Code (Código Sustantivo del Trabajo) and constitutional labor guarantees. The bill sought to complement data protection obligations rather than replace them, specifying worker‑specific duties and labor protections. Several experts and legal observatories noted overlaps and the need to harmonize definitions and competence among agencies; the proposal asked ministries to coordinate technical rulemaking to avoid regulatory gaps and duplication.

International Alignment

The draft draws on international trends and instruments (European AI Act risk‑based approach, OECD AI principles and regional initiatives) to align Colombia with evolving norms. The bill’s risk classification, human‑in‑the‑loop principle, documentation/registry expectations and emphasis on impact assessments reflect global best practices and the EU’s high‑risk model. Observers flagged the need to calibrate obligations to national administrative capacity and to ensure compatibility with trade and cross‑border data flows. See comparative commentary at Foro GPP and judicial context in recent Corte Constitucional decisions (see referenced jurisprudence summaries).

Implementation Timeline

Sources and References