Iveda launches real-time AI shoplifting detection tool

Iveda has launched a real-time zero-shot AI detection feature for its IvedaAI platform that lets users create custom video detection models with a single text prompt.

Aimed at retailers and security operators, the tool applies AI detection to live or recorded video feeds without requiring model training or data labelling. Users can enter prompts such as shoplifting, suspicious behavior, or graffiti, and the system applies a model to camera footage through the existing IvedaAI dashboard.

The launch reflects a shift in video analytics, which has typically relied on pre-built models developed through large datasets and training cycles. Iveda said its approach reduces that process to seconds by constructing a model in real time from a natural language instruction.

Retail loss prevention is one of the product's main targets. According to Iveda, shoplifting has been difficult for conventional surveillance tools because the behaviour can vary widely across store layouts, customer movements, and concealment methods.

The system uses Vision Language Models alongside Iveda's existing object detection models to identify context and intent in footage rather than only fixed objects in a frame. In early testing, prompts including suspicious behavior flagged people peering into windows and repeatedly checking staff-only entry points, the company said.

The product is being evaluated by what Iveda described as one of the world's largest fast-fashion retailers, with thousands of stores in more than 90 countries. The retailer was not named.

Retail focus

Beyond shoplifting, the prompt-based system can be used to detect loitering in sensitive areas, graffiti attempts, signs of vandalism, smoking in restricted zones, and fighting or aggressive conduct. These detections are available as soon as a prompt is submitted, according to Iveda.

The feature is available through two deployment options. In a cloud-connected mode, customers can use large language model processing for live frame analysis without adding on-site hardware. For those wanting a closed-network setup, Iveda offers an on-premise server using its Cosmos-Reason engine for local processing.

The local option is intended for retailers with stricter security or infrastructure requirements. Iveda also said it plans to ship a next-generation Cosmos-Reason-2 engine, which it expects to improve inference performance and detection accuracy.

Existing platform

The launch builds on Iveda's work in AI object detection over more than a decade, including systems for weapons recognition, licence plate identification, and facial recognition. The new feature adds Vision Language Model reasoning on top of that base for use in low-light, crowded, and complex retail settings, the company said.

David Ly, Chief Executive Officer and Founder of Iveda, described the release as a major advance in how surveillance operators can deploy AI analysis.

"This is the greatest leap forward yet in real-time AI video analytics," said David Ly, Chief Executive Officer and Founder of Iveda. "For years, building an AI detection model meant collecting data, labeling it, and waiting. Now, a retailer can type 'shoplifting' and, in seconds, have a fully functioning AI model running live across their camera network. That's never been possible before - and the accuracy of what we're detecting is remarkable."

The add-on is available now on an annual licence within the IvedaAI platform, and existing customers can enable it through their current monitoring dashboard, according to the company.

Ly also linked the launch to broader changes in AI software and hardware development.

"We've always been committed to staying at the cutting edge," Ly said. "We're already integrating the latest frameworks from NVIDIA and are continuously future-proofing the platform. 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."