AINEWS
Seeing Machines' new Physical AI Platform extends the Australian firm's driver monitoring software, deployed in more than 8 million vehicles, into factories, hospitals, and homes.
Canberra-based Seeing Machines, listed in London on the AIM growth market, has spent 25 years solving a quieter problem than the one that dominates humanoid-robotics headlines. The company's computer-vision systems already decide whether drivers are paying attention, in more than eight million vehicles on the road today. Now it wants to apply the same human-sensing stack to robots.
On 17 August 2026, Seeing Machines launched a "Physical AI Platform" that builds on its existing driver- and occupant-monitoring business. The new platform creates a real-time three-dimensional map of people, objects, and the surrounding environment, then layers in behaviour interpretation and risk anticipation. In the company's framing, it is the difference between a robot that can pick up a box and a robot that knows when the person beside it is about to step into the wrong place at the wrong time.
The pitch leans on a credential that does not show up on a typical humanoid-robotics résumé: more than eight million cars on the road already use Seeing Machines' driver- and occupant-monitoring software to detect drowsiness, distraction, and inattention. The same camera-and-vision stack that watches whether a delivery driver is nodding off is what the company now wants to put inside the next generation of humanoid robots.
"Human-Machine interaction is a complex challenge to solve," CEO Paul McGlone said in the announcement. "Our Human-Centred AI design philosophy has been solving these challenges for more than 25 years, and now we are applying it to the next frontier."
That frontier, as the company defines it, runs from factories and warehouses to hospitals, workplaces, and eventually private homes. The argument is that once robots share space with people, the binding constraint stops being dexterity or natural-language understanding, and starts being a continuous read of what people are about to do.
The launch is a press release, not a deployment. The 17 August announcement names no customers, no partners, no pilot sites, and no revenue. The eight-million-vehicle number refers to the existing automotive business, not the new platform. The small-cap Australian firm is also stepping into a category populated by better-funded competitors: Figure, Apptronik, Agility, Sanctuary, and Tesla's Optimus programme are all chasing the same humanoid-robot goal with deeper robotics engineering benches.
The Seeing Machines pitch is not about a better robot hand or a faster walking gait. It bets that human-sensing, the part of the stack that keeps a robot from hurting the people it shares a room with, is its own hard problem. The company has spent 25 years working on that problem, in cars, where the failure mode is measured in crash reports. Whether that experience translates to a hospital corridor or a factory floor is the open question the platform now has to answer.
Can the same human-sensing stack that watches eight million drivers keep a humanoid robot from hurting the people it shares a room with? The first named pilot, in a factory, a hospital, or a logistics depot, will give the first real answer.