Swann develops AI security that confronts intruders and reminds neighbours to pick up after their dogs

Homeowners must sift through pointless motion-detected alerts to get any benefit out of their security camera. A Melbourne company thinks the answer could be hurling abuse at potential thieves flagged by AI.

The system can sound a hyper-specific alert such as notifying a homeowner when a neighbour’s dog defecates on their lawn.

Home security is moving from passive monitoring to confrontational with new offerings capable of hurling profanities at burglars.

Melbourne-based Swann is developing an artificial intelligence-powered system using tools from Amazon Web Services that can be programmed to sound a hyper-specific alert to notify a homeowner when a neighbour’s dog defecates on their lawn.

The system is designed to combat relentless and pointless motion-detected alerts that can overwhelm users. The tool, called ‘Notify Me When’ allows users to input plain-text prompts to filter out irrelevant activity, focusing only on events that matter to them. It relies on AWS’s AI models to interpret video footage and generate detailed text descriptions beyond basic object recognition.

Swann chief technology officer Surjit Reghunathan said it was about enhancing amenity by solving a common bugbear.

Mr Reghunathan said that by increasing the relevance of alerts, the system has the potential to reduce unwanted notifications by up to 65 per cent. He also highlighted the feature’s customisability for more serious issues, such as ‘If a kid falls into the pool’ sought by companies in the pool industry.

“Right now on your phones will be 10 notifications from different apps, and you are busy, probably are driving, or you’re in a meeting right now. You’re not looking at it, so you have become so desensitised,” Mr Reghunathan said.

This includes using AI to act as a ‘flat mate’ to answer a doorbell when a resident is away from home. For example, a delivery person might call and the AI reply “please leave the package behind the bins”.

Surjit Reghunathan.

Mr Reghunathan said the AI could “handle the conversation for hours”.

But Swann has also developed more active security with an unreleased feature aimed at deterring break-ins. The company is developing a new functionality, referred to internally as Defender, that is triggered when a person is detected on the property and will automatically engage using human-like, aggressive language.

Users could choose a very aggressive tone and language like “Get the (expletive) out of here” in a radical departure from the recorded messages offered by competitors.

The goal is to create the unmistakeable impression that someone is home. The company is currently finetuning the system, which it plans to launch before the end of the year.

Amazon’s Ring has already incorporated extensive AI features. Swann is keen to differentiate itself with features like Swann Shield that are provided for free, unlike its competitor.

Swann, of course, is leveraging Amazon’s own foundational AI technology, specifically Amazon Bedrock and the Amazon Nova suite of models. This creates a ‘frenemy’ dynamic, where Swann is an AWS customer but a direct market competitor to an Amazon-owned brand.

Mr Reghunathan said the partnership with AWS was a pragmatic choice, leveraging the existing cloud infrastructure that already handles Swann’s media processing. The rapid iteration was also a significant advantage.

Originally published as Swann develops AI security that confronts intruders and reminds neighbours to pick up after their dogs