Critical RiskGovernment / Justice
AI Criminal Justice — No Automated Decisions
Risk assessment algorithms used in bail, sentencing, and parole decisions have been shown to exhibit racial bias and produce wildly different scores for similarly situated defendants. EU AI Act Article 5 explicitly prohibits AI systems that make final decisions in criminal justice without meaningful human review. GeraWitness provides a mandatory human-in-the-loop layer that prevents any AI risk score from being used as the sole basis for a deprivation-of-liberty decision, with full audit trail.
What GeraWitness does here
- Ensures compliance with EU AI Act Article 5 prohibitions
- Prevents discriminatory algorithmic sentencing outcomes
- Provides courts with documented human decision-making trail
- Enables meaningful challenge to AI-influenced decisions
Applicable regulations & standards
EU AI Act Article 5 (prohibited AI)European Convention on Human Rights Article 6UK Human Rights Act 1998COMPAS litigation precedents
Gera Systems integration
This use case is built into GeraCompliance. GeraWitness oversight is active by default for all high-risk actions on that platform.
Add oversight to your AI systems
GeraWitness integrates in minutes via webhook or MCP. Start with a free audit of your existing AI decision points.
Join the beta