Protent Turns Security Cameras From Passive Recordings Into Active Threat Detection

The Macro: Security Cameras Record Problems, They Do Not Prevent Them

There are over a billion surveillance cameras installed worldwide. The vast majority do exactly one thing: record footage that gets reviewed after something bad happens. A fight breaks out, someone reviews the tape. A theft occurs, someone pulls the footage. The cameras are forensic tools, not prevention tools.

The fundamental problem is attention. A single security operator might monitor 50 to 200 camera feeds simultaneously. It is physically impossible to watch all of them at once. Studies show that after just 20 minutes of watching video feeds, a human operator’s attention drops dramatically. Important events get missed because nobody was looking at the right screen at the right moment.

Motion detection and basic analytics help but produce too many false alarms. A person walking through a normal path triggers the same alert as someone behaving erratically. The noise-to-signal ratio makes operators ignore alerts, defeating the purpose.

What is needed is intelligent prioritization: AI that watches all feeds simultaneously, identifies genuinely concerning behavior patterns, and directs human attention to the situations that actually require intervention.

Protent, backed by Y Combinator, provides exactly this. Their platform detects early escalation patterns and predictive threat signals in real-time video, giving surveillance staff the context they need to intervene proactively.

The Micro: From Lockheed Martin and AWS to Predictive Security

Srihan Balaji (CEO) and Abhisheik Sharma (CTO) cofounded Protent. Srihan previously worked at Lockheed Martin Research and AWS, bringing defense and cloud infrastructure experience. Abhisheik has a background in NLP and sentiment analysis research.

The key technical claim is detecting “early escalation patterns.” This means identifying behavioral indicators that a situation is becoming dangerous before it reaches a critical point. Aggressive body language, unusual crowd movements, individuals acting erratically, and other pre-incident indicators that trained security professionals recognize but cannot monitor across 200 feeds simultaneously.

This is technically challenging. Distinguishing between a heated conversation and a pre-fight escalation requires understanding context, body language, spatial relationships, and temporal patterns. False positives that send security to investigate normal interactions waste time and erode trust. False negatives that miss genuine threats have serious consequences.

Competitors include Verkada (smart cameras with cloud management), Milestone (video management software), and BriefCam (video analytics). Most of these focus on post-event search and basic detection rather than predictive escalation monitoring.

The GovTech angle is significant. Government facilities, public spaces, transit systems, and critical infrastructure all need security monitoring at scale. Government procurement cycles are slow but contracts tend to be large and long-term.

The Verdict

Protent is tackling a high-value problem with clear demand. Proactive security monitoring is better than reactive footage review, and AI is the only way to achieve it at scale.

At 30 days: what is the detection accuracy for genuine escalation events, and what is the false positive rate?

At 60 days: are security operators trusting and acting on Protent’s alerts, or are they ignoring them?

At 90 days: has Protent prevented a real incident through early detection that would have been missed by human monitoring?

I think Protent is building the right product for a large market. The gap between what security cameras could do and what they actually do is enormous. AI that intelligently prioritizes human attention across hundreds of feeds is the next evolution of physical security. The defense industry background of the founding team lends credibility for government and enterprise sales.