AI Model Trained on Prison Calls to Predict Crimes

The Initiative Behind AI in Correctional Facilities

A US telecom company has developed an AI model trained on years of phone and video calls made by inmates. The goal is to scan these communications in hopes of predicting and preventing crimes.

Foundation of the AI Model

Securus Technologies' president, Kevin Elder, shared that the company started this initiative in 2023. With a significant database of recorded calls, they trained AI models intended to detect criminal activities. One notable model originated from analyzing seven years of calls made within Texas prisons, and they're also developing state- or county-specific versions.

Real-Time Monitoring Implementation

Over the last year, Securus has been piloting these AI tools to monitor inmate conversations in real-time. While the precise locations remain undisclosed, their clients include various detention centers, prisons, and ICE facilities.

Early Detection of Criminal Intent

Elder emphasized, “If we can point that large language model at an entire treasure trove of data, we can detect and understand when crimes are being thought about or contemplated.” This proactive approach helps catch potential criminal activities early in the planning phase.

Operational Mechanism

According to Elder, Securus' monitoring tools work by analyzing conversations either at random or focusing on individuals flagged for suspicious activities. The AI analyzes calls, texts, and emails, flagging sections for human review, which can then be escalated to investigators.

Impact on Crime and Safety

Elder noted that these AI features have successfully disrupted human trafficking and gang operations orchestrated from within the prison system. However, specific cases uncovered by the new AI models were not disclosed to MIT Technology Review.

Ethical Concerns About Surveillance

While inmates and those they call are made aware that conversations are recorded, there’s ongoing debate about the implications of using these recordings to train AI models. Bianca Tylek, from the prison advocacy group Worth Rises, argued that this is an example of coercive consent, pointing out that inmates pay for these calls while also having their data exploited.

Monitoring Policies and Imposed Constraints

Securus operates under the recording and monitoring policies dictated by correctional facilities, which raises questions about opt-out options for inmates regarding AI training.

Privacy Violations and Surveillance Criticism

Advocates for inmate rights, including the ACLU’s Corene Kendrick, express concern about the invasive nature of such monitoring technologies. Kendrick stated that monitoring inmates’ every conversation raises serious civil liberty issues, warning that the technology may outpace legal frameworks designed to regulate it.

Funding and Regulatory Landscape

Securus recently gained traction with regulators regarding telecom fund allocation, which permits the company to finance its AI tools. 2024 FCC reforms prohibited passing monitoring costs to inmates, compelling correctional facilities to cover these expenses, which caused backlash among law enforcement agencies.

Legislative Changes and Responses

As Securus worked on its AI tool, it sought to influence FCC regulations, prompting discussions about funding surveillance technology with fees collected from inmate calls. The FCC has since reconsidered these regulations in favor of telecommunications entities.

The Future of AI in Corrections

With recent FCC rulings enabling the passing of security cost burdens back onto inmates, Securus reaffirms that such tools are essential for safety in correctional environments, despite ongoing dissent from civil liberties advocates.