The Government's plan to regulate AI in Colombia: the main points of the bill that seeks to regulate this technology

Following the example of countries in the region such as Argentina, Brazil and Chile, and the pioneering step taken by the European Union in 2024, when it adopted the world's first legal standard on artificial intelligence , in Colombia, the National Government, headed by the Ministry of Science, Technology and Innovation, resumed efforts for this legislature to get a bill approved that seeks to regulate the development and use of this technology in Colombia .

The initiative, which had already foundered in June after failing to achieve the necessary quorum to discuss the bill in the Sixth Committee, was resubmitted this week for consideration in Congress. According to Science Minister Yesenia Olaya, the new text is "a consensus built between academia, the productive sector, the private sector, and more than thirteen previous proposals."

These are the main points included in the proposal, which aims to "enable ethical, responsible, competitive, and innovative adoption of AI systems, incorporating technical, human rights, and productive development criteria," as well as some of the changes this project brings compared to its previous version.

The bill establishes that the Ministry of Science will be the competent national authority to guide the implementation of this law and coordinate AI governance in Colombia. Among the ministry's responsibilities will be the adoption of guidelines, technical standards, and best practices based on scientific evidence, national and international regulatory developments, and the principles established in the law .

"The Ministry will perform technical, advisory, and coordination functions, promoting collaboration with other public entities, the productive sector, academia, regional governments, and civil society, and fostering mechanisms for self-regulation, certification, and responsible evaluation by stakeholders in the AI ecosystem," the proposal states.

In the case of high-risk AI systems, the Ministry of Sciences may issue technical recommendations or establish additional requirements to mitigate significant risks that affect fundamental rights, security, or collective well-being. This is a change from the previous draft, which, in the event of an unacceptable risk to fundamental rights, proposed the Ministry's ability to recommend or order, directly or through the competent authorities, the suspension, blocking, deactivation, or cessation of operation of these AI systems .

As part of the measures to promote the development of AI in the country, the government's proposal also considers the creation, operation, and evaluation of regulatory sandboxes for artificial intelligence. These are defined in the bill as "controlled and flexible testing environments that allow for experimentation, adaptation, and scaling of AI solutions under safe, ethical, and supervised conditions, even when they pose potential risks to fundamental rights or the public interest."

"Each approved sandbox must have a specific plan that includes objectives, scope, participating actors, technical and legal conditions, proportional risk mitigation measures, monitoring mechanisms, and evaluation criteria, ensuring the traceability of the lessons learned. This plan will be binding on those who participate in the test," the bill states.

A prioritization criterion is also being considered for science, technology, and innovation projects seeking access to tax benefits for developing AI-based solutions or adopting these technologies to enhance their production processes.

The government's proposal considers the design and implementation of AI education, training, and capacity building programs to strengthen skills and competencies at all levels and modalities of education , including early childhood, primary, secondary, technical, technological, higher, and adult education. Furthermore, they aim to ensure ongoing training in this field for school principals and teachers.

In terms of education, the bill also proposes including content related to data science, computational thinking, digital ethics and artificial intelligence at the basic and secondary education levels, the development of curricula oriented to practical applications, interoperability, sustainability and ethics of automated systems with the Sena, the creation of academic programs, faculties or departments specialized in artificial intelligence in higher education institutions, as well as the creation of common cores or elective subjects for general training that are transversal to careers in different areas of knowledge .

Another important point raised in the bill is aimed at a fair transition in response to the impact of AI on employment. In this regard, it states that " employers who implement AI systems that could generate displacement or substantial transformation of existing job functions must adopt retraining, relocation, or training plans for potentially affected workers ." This will be overseen by the Ministry of Labor.

If the government's proposal is approved, it will also lead to the recognition of new forms of crime related to artificial intelligence. This task will fall to the Ministry of Justice, in coordination with the High Council for Criminal Policy, entities that must submit a regulatory proposal to Congress to update the Criminal Code to criminalize conduct that, "through the improper, malicious, or fraudulent use of Artificial Intelligence systems, generates serious negative impacts on fundamental rights, the integrity of individuals, or the public interest ."

Among the modifications brought about by this new proposal, developed by the Ministry of Sciences in coordination with the National Planning Department, the Ministry of Information and Communications Technology, and the Ministry of Education, is a new risk classification for artificial intelligence. What were previously referred to as prohibited AI systems are now classified as "High Risk."

These are "those that, by their nature, context, and intended purpose, present a high risk of adverse impact on the health, safety, or fundamental rights of individuals." According to the bill, these systems will be subject to specific requirements for conformity assessment, risk management, transparency, and human oversight prior to marketing, commissioning, and upgrades.

Additionally, systems that were previously classified as limited risk are considered in the new proposal as AI systems with specific transparency obligations . These are systems that, although not classified as high risk, must comply with specific transparency obligations to ensure that users are properly informed, such as those intended to interact directly with people or those that generate or manipulate image, audio or video content to make them appear authentic.

Finally, the new draft includes minimal or zero-risk AI systems, a classification that includes systems that do not fall into the previous categories. According to the draft, the voluntary adoption of codes of conduct and best practices will be encouraged.