California-based Verkada is using AI to detect potential intruders and deter them with a customised message based on what they are wearing.

This can proactively helps prevent crime by detecting and warning potential intruders before raising an alarm. This innovation addresses a gap in traditional security systems, which only react after a break-in has occurred.

Leveraging large vision, audio and language models, it analyses footage from video security cameras to detect unauthorised behaviour such as loitering and automatically triggers a sequence of AI-generated messages that incorporate scene context, such as describing a person’s clothing or location, and escalate deterrence measures over time if the person remains on site. The system can also change the voice for each deterrence incident for greater effectiveness against repeat offenders.

“AI-powered deterrence represents a fundamental shift that we’ve been driving for our customers: moving from reactive to proactive security,” said Brandon Davito, senior vice president at Verkada. “We’re delivering on Verkada’s goal of building safer, smarter communities in a very real and tangible way when our customers have the ability to stop an incident before it escalates. This impact will only scale as we continue to invest in bringing even more AI-powered capabilities into Verkada Command.”

Organisations can configure AI-powered deterrence (www.verkada.com/blog/ai-deterrence-enhancements/) to create intelligent workflows that begin with detection and escalate only when needed. Here is an example of how the system might respond to a person loitering outside of a property after-hours:

• Initial warning upon detection of the person loitering: “A reminder that this is private property. Please come back during business hours.”
• If loitering continues, a context-aware message: “Hey you, with the red hoodie and backpack, you can’t be here at this time. Please leave now before our alarms are triggered.”
• If loitering continues, a final warning: “Hey, this is your final warning. If you don’t leave now, I’ll have to take immediate action.”
• If loitering continues, an alarm is raised and Verkada’s system can be configured to activate a siren or strobe, initiate live agent talk-down, notify a contact list of internal security teams, or dispatch the police.

This approach enables, for example, car dealerships to deter catalytic converter thieves, schools to prevent graffiti vandalism, and businesses to protect outdoor assets, all without unnecessary alarm activations.

Verkada also announced the availability of more advanced AI-powered alerting capabilities, including compound alerts, which enable users to trigger notifications only when multiple conditions they specify are met. For example, a user could ask Verkada Command to send them an alert when a person crosses a restricted line and is not wearing a hard hat. Or they could choose to set up loitering alerts only for individuals who have been identified as a person of interest.

All Verkada’s intelligent alerts can now be pushed seamlessly to external applications outside of Verkada Command, such as PowerBI and Tableau, through newly available webhooks.

To speed up investigations and reporting, AI-powered incident summaries automatically generate descriptions for video archives that are saved within an incident report.

More than 30,000 organisations across 171 countries worldwide use Verkada (www.verkada.com) as their physical security layer for easier management, intelligent control and scalable deployments.