A US telecom company is testing AI prison call monitoring tools that aim to identify criminal activity earlier by scanning communications from incarcerated people. The system, developed by Securus Technologies, is built from years of recorded inmate calls and is now being piloted across phone and video calls, text messages, and emails.
The project has become a flashpoint because it sits at the intersection of prison surveillance, artificial intelligence, family communication, and call fees. Securus says the technology can help facilities review communications more efficiently and identify threats. Prison rights advocates argue that the data behind the system was collected in a setting where meaningful consent is difficult, if not impossible.
How Securus built the AI model
Securus Technologies president Kevin Elder told MIT Technology Review that the company began building its AI tools in 2023. The company used its large archive of recorded inmate communications to train models designed to detect criminal activity.
One model was created using seven years of calls made by inmates in the Texas prison system. Securus has also been working on state- or county-specific models, according to Elder.
The purpose is not only to find explicit discussion of a crime after the fact. Elder described a more predictive ambition: using a large language model to review a broad body of communications and identify when criminal activity appears to be forming.
“We can point that large language model at an entire treasure trove [of data],” Elder says, “to detect and understand when crimes are being thought about or contemplated, so that you’re catching it much earlier in the cycle.”
Over the past year, Elder says, Securus has been piloting these AI tools to monitor inmate conversations in real time. The company declined to say where the pilots are taking place.
What the system scans and flags
According to Elder, investigators at detention facilities can use the AI features in ways similar to other monitoring tools. They may monitor randomly selected conversations or communications involving people suspected by facility investigators of criminal activity.
The model can analyze several types of communication:
- phone calls
- video calls
- text messages
- emails
When the AI identifies sections that require attention, those sections are flagged for human agents to review. The agents can then send material to investigators for follow-up.
Elder said Securus’ broader monitoring work has helped disrupt human trafficking and gang activities organized from within prisons, and has also been used to identify prison staff bringing in contraband. However, the company did not provide MIT Technology Review with any cases specifically uncovered by its new AI models.
Securus’ customers include jails holding people awaiting trial and prisons for people serving sentences. Some facilities using Securus technology also have agreements with Immigrations and Customs Enforcement to detain immigrants, though Securus does not contract with ICE directly.
The consent problem
People in prison, and the people they call, are notified that conversations are recorded. The dispute is whether that notification is enough when recordings may also be used to train an AI model.
Bianca Tylek, executive director of the prison rights advocacy group Worth Rises, told MIT Technology Review that awareness of recording does not mean people understand or accept the use of their communications as AI training data.
“That’s coercive consent; there’s literally no other way you can communicate with your family,” Tylek says.
Tylek added that because inmates in the vast majority of states pay for these calls, they are not being compensated for their data and are instead paying while that data is collected.
A Securus spokesperson said the use of data to train the tool "is not focused on surveilling or targeting specific individuals, but rather on identifying broader patterns, anomalies, and unlawful behaviors across the entire communication system." The spokesperson said correctional facilities set their own recording and monitoring policies, which Securus follows, and did not directly answer whether inmates can opt out of having recordings used to train AI.
Civil liberties concerns
Advocates also point to Securus’ prior record. Leaks of the company’s recording databases showed that it had improperly recorded thousands of calls between inmates and their attorneys.
Corene Kendrick, the deputy director of the ACLU’s National Prison Project, said the new AI system expands an already invasive surveillance environment. She also said courts have set few limits on this kind of power.
“[Are we] going to stop crime before it happens because we’re monitoring every utterance and thought of incarcerated people?” Kendrick says. “I think this is one of many situations where the technology is way far ahead of the law.”
The company spokesperson said the tool is meant to make monitoring more efficient amid staffing shortages, “not to surveil individuals without cause.”
Who pays for AI surveillance
The future of the tool is also tied to how inmate call fees can be used. In 2024, the Federal Communications Commission issued a reform that barred telecom companies from passing the costs of recording and surveilling calls on to inmates. Companies could still charge inmates a capped rate for calls, but prisons and jails were told to pay most security costs from their own budgets.
The change drew quick opposition. Associations of sheriffs said they could no longer afford proper call monitoring, attorneys general from 14 states sued over the ruling, and some prisons and jails warned that they would cut off access to phone calls.
While building and piloting its AI tool, Securus met with the FCC and lobbied for a rule change. The company argued that the 2024 reform went too far and asked the agency to again allow companies to use fees collected from inmates to pay for security.
In June, Brenda n Carr, whom President D onald Trump appointed to lead the FCC, said the agency would postpone all deadlines for jails and prisons to adopt the 2024 reforms. Carr also signaled support for helping telecom companies fund AI surveillance with fees paid by inmates. In a press release, Carr wrote that rolling back the reforms would “lead to broader adoption of beneficial public safety tools that include advanced AI and machine learning.”
On October 28, the FCC voted to pass new, higher rate caps and allow companies such as Securus to pass security costs related to recording and monitoring calls on to inmates. Those costs can include storing recordings, transcribing them, or building AI tools to analyze calls.
A Securus spokesperson told MIT Technology Review that the company aims to balance affordability with funding safety and security tools. “These tools, which include our advanced monitoring and AI capabilities, are fundamental to maintaining secure facilities for incarcerated individuals and correctional staff and to protecting the public,” the spokesperson wrote.
FCC commissioner Anna Gomez dissented. “Law enforcement,” she wrote in a statement, “should foot the bill for unrelated security and safety costs, not the families of incarcerated people.”
The FCC will seek comment on the new rules before they take final effect. Until then, the debate around AI prison call monitoring remains centered on three unresolved questions: how inmate data should be used, how much surveillance is justified, and whether incarcerated people and their families should pay for tools built to monitor them.