Iveda Adds Real-Time Zero-Shot AI Detection to Video Analytics Platform

Iveda, a provider of AI-driven video analytics and smart city technologies, has announced a new capability within its IvedaAI platform that enables real-time zero-shot AI detection using natural language prompts.

The feature allows retailers and security operators to enter a word or phrase, such as "shoplifting," "suspicious behavior," or "graffiti," and immediately activate a custom AI detection model that begins analyzing live video feeds. According to the company, the process requires no model training, data labeling or deployment delays.

Iveda said the new functionality represents a significant change in AI-powered video surveillance workflows. Rather than relying on extensive datasets and lengthy model training processes, the platform creates a custom AI model in real time from a user-submitted prompt and applies it to live or recorded video streams within seconds.

"This is the greatest leap forward yet in real-time AI video analytics," said David Ly, CEO and founder of Iveda.

The company highlighted the technology's potential for retail loss prevention, an area where traditional AI surveillance systems have faced challenges due to the wide variety of shoplifting methods and retail environments. Iveda said its zero-shot approach combines advanced Vision Language Models (VLMs) with the company's existing object detection models to analyze behavior, intent and context rather than focusing solely on objects appearing in a video frame.

According to Iveda, the technology is currently being evaluated by a major fast-fashion retailer with thousands of store locations operating in more than 90 countries. The company said early testing demonstrated the system's ability to identify behaviors such as individuals peering into windows and repeatedly monitoring staff-only access points after users entered prompts related to suspicious activity.

In addition to shoplifting detection, the platform can be prompted to identify suspicious or pre-incident behavior, loitering in sensitive locations, graffiti attempts, vandalism indicators, smoking in restricted areas and fighting or aggressive conduct. These detections are activated directly through the existing IvedaAI dashboard without requiring new infrastructure or separate logins.

To support different deployment requirements, Iveda is offering the technology through two deployment options. The first is a cloud-connected mode that uses large language model processing for live frame analysis without additional on-site hardware. The second is a local deployment option powered by the company's Cosmos-Reason engine, which operates on a dedicated on-premises server. Iveda said this configuration provides fully local processing with no cloud dependency and no open ports.

The company also announced plans to begin shipping its next-generation Cosmos-Reason-2 engine within the coming quarter, which is intended to further improve inference performance and detection accuracy.

Iveda said the new capability builds on more than a decade of AI development that includes object detection technologies for weapons recognition, license plate identification and facial recognition. The company noted that the addition of Vision Language Model reasoning is intended to improve performance in low-light, crowded and complex retail environments.

"We've always been committed to staying at the cutting edge," Ly said. "What we're announcing today is the result of years of foundational work meeting this new wave of AI capability — and retailers are going to feel the difference immediately."

Iveda said real-time zero-shot detection is available immediately as an add-on annual license for the IvedaAI platform, and existing customers can activate the feature through their current secure camera monitoring dashboard.