BlackBerry says robotics is running into a code problem, not a metal one
Physical AI increasingly depends on the software layer that keeps the robot under control.📷 AI-generated image / TECH&SPACE
- ★QNX research puts software and security at the center of barriers to physical AI robotics.
- ★The problem grows as robots enter less controlled, more open and operationally risky environments.
- ★For robotics companies, this raises the importance of architecture, cybersecurity and validation, not only sensors and actuators.
The Robot Report reports QNX research with a blunt message for robotics: in physical AI, software is increasingly becoming the main bottleneck. That changes how progress in robotics should be read. For years, attention naturally went to arms, wheels, joints, batteries, cameras and lidars. Now the pressure is moving toward the code that has to connect perception, decision-making, safety and response in the real world.
The context is simple, but uncomfortable. Robots are no longer imagined only as machines sealed inside predictable cells or clearly marked industrial zones. As they move into less constrained environments, the software layer has to absorb much more mess: changing objects, nearby people, unknown edge cases, communication failures and security threats. That is why a finding that software and security are becoming higher priorities is not a bureaucratic footnote. It is a signal that physical AI cannot scale through a faster model or a stronger actuator alone.
QNX is a relevant name here because it is associated with embedded and safety-sensitive systems, the kind of infrastructure where a failure is not just an app crash. In robotics, the stakes are sharper: a bad decision can move a real machine, stop a line, damage goods or create risk for people. So the software question in physical AI is also a question of architecture, validation, cybersecurity and accountability.
QNX research, reported by The Robot Report, puts software and security ahead of pure mechanics as robots move into less controlled spaces.
Security, validation and embedded systems are becoming core parts of the robotics product.📷 AI-generated image / TECH&SPACE
It is useful to separate the hype from the operating reality. Physical AI sounds like a natural extension of the generative AI wave, but a robot is not a chatbot with motors. It has to work in time, space and safety constraints. It has to know when it does not know enough. It needs a predictable route to a safe state. It needs updates that do not open new weaknesses. And it needs to explain at least part of its behavior to the people maintaining, certifying or deploying it.
That is why links to The Robot Report’s robotics coverage and QNX’s work around embedded software are useful context for this finding. The story is not just one survey. It fits a broader industrial pattern: hardware keeps improving, AI models are more accessible, but system integration remains expensive, slow and fragile. If software cannot reliably orchestrate sensors, models, controllers and safety policies, the robot remains a demo rather than a product.
For robotics companies and buyers, the practical consequence is clear. The next phase of physical AI will not be won only by adding more parameters, more cameras or more marketing language. It will be won by teams that can prove their software layer works under pressure, is protected, can be maintained and behaves predictably enough when the world stops looking like a lab.
