This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
In 2018, the Salta provincial government in Argentina, in partnership with Microsoft, deployed an AI system to predict teenage pregnancies among low-income and marginalized populations. The system utilized invasive personal data, including home infrastructure and disability status, and was criticized by activists and researchers for methodological flaws, lack of transparency, and potential for discriminatory surveillance and eugenic-style population control.
In 2018, during the abortion-decriminalization debate in Argentina, the Salta city government deployed a teenage-pregnancy predictive algorithm built by Microsoft that allegedly lacked a defined purpose, explicitly considered sensitive information such as disability and whether their home had access to hot water.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Unequal treatment of individuals or groups by AI, often based on race, gender, or other sensitive characteristics, resulting in unfair outcomes and unfair representation of those groups.
Also classified as
The system relied on highly invasive data collection, including GPS locations, photos, and personal details of 12,000 young girls.
The system targeted low-income, migrant, and Indigenous girls, unfairly labeling them as 'predestined' for teenage pregnancy based on biased socioeconomic variables, which represents unfair discrimination and misrepresentation.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The risk and resulting harms were caused by the intentional decision of human actors in the Salta government and Microsoft to deploy an invasive, unregulated predictive surveillance system targeting vulnerable populations.
Added categories peak at 3: Substantial (Epistemic Harm, Indirect)
The Salta provincial government in Argentina deployed a highly flawed AI system developed with Microsoft to predict teenage pregnancy. The system targeted 12,000 marginalized and Indigenous girls, conducting invasive surveillance and violating fundamental privacy and human rights. While it did not threaten physical infrastructure or national sovereignty, it constitutes a substantial abuse of technology for state-sponsored algorithmic discrimination.
The system operated as a predictive tool, taking specific demographic and socioeconomic inputs to output a pregnancy risk score without maintaining an active dialogue or executing autonomous real-world actions.
Unacceptable Risk: The system is an exploitative tool targeting vulnerable populations based on age, disability, and socioeconomic status for state surveillance and population control, which violates fundamental rights.
The system was designed to model and predict the future behavior and life events (specifically teenage pregnancy) of young girls based on demographic and environmental data.
The reports provide consistent, detailed accounts of the system's deployment, the specific data collected, the target demographic, and the technical flaws identified by the University of Buenos Aires. The technical analysis of the SMOTE data contamination is highly specific and credible.
The developers framed the system as a pioneering public health tool to prevent teenage pregnancy, whereas critics interpreted it as an authoritarian surveillance and population control mechanism.
Technology
Data Contamination via SMOTE
SMOTE applied before data split artificially inflated the accuracy.
Methodological Design Errors
Used current pregnancy status to predict future pregnancy years later.
Opaque Algorithmic Models
The social theories and algorithms used in the system were completely opaque.
Data Inputs
Incomplete and Biased Data
Sensitive data was hidden, leading to biased socioeconomic representations.
Missing Critical Variables
Excluded key public health variables like sex education and contraception.
Invasive Data Collection
Agents gathered photos, GPS locations, and sensitive surveys from minors.
Human Factors
Overreliance on AI Predictions
Officials treated flawed predictions as absolute, predestined truths.
Power Asymmetry and Coercion
Vulnerable families could not refuse due to dependency on state aid.
Political and Ideological Bias
Anti-abortion groups used the tool to argue against reproductive rights.
Process and Methods
No Independent Validation
The system lacked external technical validation before public deployment.
Lack of Transparency
No official reports or data on accuracy were ever published.
Poor Machine Learning Practices
Basic ML principles were violated by mixing training and evaluation data.
Management
Prioritizing PR over Safety
Officials promoted the tool on TV before verifying its technical validity.
Lack of Accountability
Government and Microsoft failed to address criticisms or publish outcomes.
Inadequate Risk Assessment
Management failed to evaluate ethical and social impacts on young girls.
Regulatory Environment
Absence of AI Regulation
No national laws existed to regulate or restrict AI system deployments.
Lack of Oversight and Audits
No government body assessed the system's impact on girls and women.
No Human Rights Framework
Lacked legal protections for vulnerable groups against algorithmic abuse.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
