LEO Technologies extends prison surveillance from calls to voices, faces, and behavior

LEO Technologies has launched a pair of AI products that expand correctional surveillance beyond the content of incarcerated people’s communications, allowing agencies to identify speakers through voice biometrics and continuously analyze camera feeds for movements, relationships, and behavior the company says may signal an emerging threat.

Verus Voice AI is designed to detect when someone uses another incarcerated person’s personal identification number to place a call.

The launches represent a significant broadening of correctional monitoring. Instead of requiring an investigator to search calls or review recorded video after an incident, LEO Technologies is marketing Verus as a continuously operating intelligence system that can connect a person’s communications with their physical movements and institutional history.

Voice AI addresses a longstanding weakness in prison telephone systems. Calls are generally attributed to the PIN entered before they are placed, but PINs and telephone access can be shared, traded, coerced, or stolen. That can allow one person to use another’s account, obscuring who is communicating with people outside the facility.

Verus Voice AI analyzes the speaker rather than relying on the account or device.

The company says its agentic AI continuously develops voice identities across monitored calls and alerts investigators when the detected speaker does not appear to match the person assigned to the PIN.

The system is intended to identify possible PIN sharing, identity fraud, and unauthorized communications across large quantities of recorded call data.

The product can also be used as an investigative voice seeker whose name or account information is unknown.

The company describes the capability as a way to “surface hidden voices” and connect individuals across different conversations.

A suspected mismatch can then be cross-referenced through the larger Verus platform against the person’s previous calls and facility history.

LEO Technologies says the platform ingests communications from existing telephone and messaging systems, uses Amazon Transcribe to produce text and applies semantic, generative, and agentic AI to identify context, coded language, relationships and possible threats.

The company says Voice AI findings must be reviewed by qualified personnel before action is taken.

That safeguard is significant because a biometric match could affect disciplinary investigations, housing decisions or referrals for criminal investigation.

Verus Vision AI moves the same continuous monitoring model into correctional video systems. Rather than requiring facilities to replace their cameras, the product is designed to analyze feeds from infrastructure they already own, both in real time and retrospectively.

It applies facial recognition, gait analysis, and behavioral analytics to existing correctional camera systems. Together, the products extend the company’s Verus platform across communications, biometric, and video data.

The company says the system can detect crowd formation, fighting, falls, isolation, unauthorized movement, perimeter activity, and environmental changes.

When an incarcerated person is missing during an automated count, the AI can initiate a search across camera feeds. It can also track movement patterns and interactions over time to identify relationships, group dynamics, and activity that investigators may consider suspicious.

LEO Technologies says Vision AI can use facial recognition and gait to identify individuals involved in an incident.

The company also claims the system can recognize facial expressions indicating that an attack may be imminent, and detect visual signs associated with suicidal ideation, severe depression, or self-harm.

Those claims move the system beyond identifying objects or documenting movement and into interpreting intent and psychological condition from observable behavior.

Vision AI also develops what the company calls long-term behavioral intelligence through social network analysis.

By examining who spends time together, where meetings occur, and how movement patterns change, the system is intended to help correctional officials map affiliations and identify evolving institutional risks.

LEO Technologies says the video platform can support compliance with the Prison Rape Elimination Act by detecting circumstances that may warrant scrutiny, including multiple incarcerated people entering one cell or staff and prisoners appearing together in unauthorized locations.

The system is intended to preserve video evidence and alert personnel before a suspected situation develops into an assault.

The technology is also being marketed as a response to correctional staffing shortages. Inmate counts can occur between five and 11 times each day, depending on the facility and security level.

LEO Technologies says Vision AI can continuously verify locations and reduce the staff time spent conducting manual counts, leaving officers available for medical escorts, programming, rehabilitation, and direct supervision.

The company’s product page estimates that correctional institutions spend more than $1.5 billion and 62.7 million staff hours annually on counting and monitoring, and projects that automation could recover 37.6 million hours and repurpose more than $627 million in labor.

The two products fit within a much larger effort to create a unified correctional intelligence environment.

LEO Technologies says the Verus suite can combine communications analysis, voice biometrics, video, incident reports, and other institutional records so that an alert in one system can be evaluated alongside information held elsewhere.

LEO Technologies says human review remains part of both products and that the systems deliver recommendations rather than autonomous disciplinary decisions.

Article Topics

biometric monitoring  |  biometrics  |  facial recognition  |  gait recognition  |  prisons  |  video surveillance  |  voice biometrics