September 18, 2026
Most days we use this space to dig into humanoid robots specifically -- machines with a torso, two arms, and two legs that are meant to work alongside people. Today's two headlines don't really qualify. One is about cybersecurity for AI-driven robots and industrial systems broadly. The other is about the environmental cost-benefit math of building robots at all. Neither names a bipedal machine, a walking gait, or a two-handed manipulation task. So rather than force a humanoid angle onto stories that aren't about humanoids, we're going to be straight with you about what these pieces actually say -- because the underlying arguments still matter to anyone evaluating automation, humanoid or otherwise.
IEEE Spectrum's piece, sponsored by VicOne, makes a point that's easy to miss if you're only thinking about mechanical failure: the old robot-safety question was whether a machine stays safe when something goes wrong internally. The new question is whether a machine stays safe when an outside attacker deliberately corrupts what it perceives, decides, or does -- with nothing visibly broken. That's a meaningfully different threat model. A robot that reads a manipulated camera feed or a poisoned sensor input doesn't look damaged. It looks confident, and it's wrong. We've made this exact argument before about humanoid robots specifically: the real vulnerability in a robot that can see, decide, and act on its own isn't the gears, it's the perception-to-action pipeline. What's notable here is that the same logic now applies to any AI-driven physical system -- industrial arms, autonomous vehicles, warehouse robots. If your business is evaluating any physical-AI deployment, the question to ask a vendor isn't "what happens if this breaks" but "what happens if someone feeds it a lie it believes." Most procurement checklists still don't ask that.
Robohub's piece poses a simple but underrated question: if a robot costs more, materially and energetically, than the benefit it delivers, is it actually a win? The example given is a solar-panel installation robot -- faster and more precise than a human crew, but built from expensive materials and drawing down a heavy battery every few hours. That's not a knock on the robot's capability. It's a challenge to anyone selling automation on speed and precision alone, without accounting for the full lifecycle cost. This is the conversation humanoid robotics companies are going to have to have eventually, and mostly haven't yet. A humanoid platform marketed as a general-purpose worker is, by definition, more materially and energetically expensive than a purpose-built machine that does one job. Robohub's framing suggests that as buyers get more sophisticated, "it can do everything" stops being a selling point on its own and starts being a cost question: everything, at what embodied cost? We think that's the right instinct, and it's one more reason we've argued that the real story in humanoid robotics right now isn't the acrobatics -- see our take on Digit 5's safety bet -- it's whether the economics and safety case actually hold up under normal, boring, everyday use.
Neither story announces a new product you can buy today, and that's fine -- these are the questions vendors will be graded on next. If you're building or budgeting for any AI-driven automation, robotic or software, the same two questions apply: what's the attack surface when the system's perception can be manipulated, and what's the true lifecycle cost versus the benefit it delivers. Those aren't robotics-specific questions either. They're the same due diligence we'd push on any AI agent making decisions inside your business, which is part of why security and workflow transparency matter as much for software agents as they will for humanoid ones.
So here's the honest question for you: when a robotics headline doesn't actually deliver the humanoid news you were hoping for, is a broader cybersecurity or sustainability lens like this still useful to you -- or would you rather we just tell you plainly that it was a quiet day for humanoids and stop there?
Sources