Greece - Eva Algorithm for Border Control in the Pandemic

What Happened

During the COVID-19 pandemic, the Greek government deployed an AI algorithm named "Eva" to manage border control and optimize the allocation of limited testing resources. Eva analyzed traveler data (such as country of origin, age, gender, and travel history) to predict the likelihood of a traveler testing positive, selectively directing high-risk travelers for testing. While celebrated for its resource efficiency, the system was criticized for systemic bias, opaque decision-making, and potential discrimination against travelers from certain regions. Some researchers argued the model was a "black box" that lacked independent, transparent validation, creating a precedent of algorithmic profiling at international borders.

How MEG 1 would have acted

Eva's opaque decision-making process violated the requirements of informational transparency and non-discrimination. Under MEG 1, border control algorithms must undergo continuous, independent Dynamic Accuracy Index (DAI) audits to verify they do not rely on biased demographic proxies.

MEG 2 Analysis - Legal Framework

Agent level: N2 (management) - automated risk-profiling system used by a public authority to make decisions affecting the freedom of movement and public health status of travelers.
Cause (6.1):
(a) system defect - the lack of independent validation and potential demographic bias built into the predictive model (design defect).
(b) autonomous decision error - the system's decision to flag specific travelers for testing based on opaque, algorithmic risk scores.
Attachment of liability: The Greek government as the operator is liable for deploying an unvalidated, opaque profiling system; the developer is liable for the model's design defects (6.1a) regarding transparency.

Procedural Mechanism

6.2 (architectural human confirmation): Coercive administrative actions (such as directing a traveler for mandatory medical testing or border detention) must require architectural human confirmation. The AI can only suggest risk levels; a human officer must verify the recommendation.
7.1 (independent forensics): The specific traveler data, the risk indicators used, and Eva's recommendation must be securely logged in an unalterable ledger to allow for human rights and judicial audits.
9.4 (least privilege): The profiling system was granted broad authority over traveler processing without public validation or independent oversight layers, violating safe operational boundaries.