Humanoid face robots have reached a point where they can move their lips in sync with speech and display detailed facial expressions. The new arXiv preprint 'RoboFake' points out that if such a robot is given a real person's face and driven by generated speech, it becomes a face forgery that happens in the physical world rather than on a screen.

Most deepfake detectors are trained on digital manipulations, so the paper asks whether they are ready for this kind of robotic forgery. The abstract does not report results, but the framing itself highlights a potential blind spot: a convincing fake face that is not a video edit but a physical object.

Because the source is a single preprint with a truncated abstract, this digest can only describe the question posed, not any findings. The significance is that the definition of a deepfake may need to expand to include physical robot faces.