Repeated failures of electronic ankle monitors lead to AI-driven behavioral intent prediction systems that restrict the physical movement of former convicts in real time.
After a string of horrific crimes committed by ankle-monitored offenders who conducted meticulous reconnaissance of their victims, public outrage overwhelms civil liberties arguments. South Korea pilots an AI system that doesn't just track where offenders are, but predicts where they intend to go and why. The system works — recidivism drops 60 percent in the first year. But the technology doesn't stay confined to convicted criminals. Insurance companies demand access to threat scores. Landlords use them to screen tenants. Employers require clearance certificates. Within five years, a permanent underclass emerges — people whose algorithmic profiles ensure they can never fully re-enter society, regardless of their legal status.
Choi Dongwook, 41, released from prison two years ago for a non-violent fraud conviction, stands frozen outside a convenience store in Suwon at 11 PM on a Wednesday. His phone has just vibrated with an AEGIS alert: his threat score spiked because he deviated from his usual route home and entered a residential zone where a registered victim lives — someone he has never met and whose existence he was unaware of. He has nine minutes to return to his approved corridor before his parole officer is automatically notified. He turns around and walks back the way he came, past the store where he was going to buy milk for his daughter.
Predictive systems are only as good as their training data, which is saturated with the biases of the criminal justice system that produced it. AEGIS would disproportionately flag low-income and minority individuals whose daily patterns — irregular work hours, frequent address changes, cash transactions — algorithmically resemble 'threat indicators.' The system would not prevent crime so much as automate discrimination, creating a technologically enforced underclass that mirrors existing social inequalities.