I'm taking yevoc's measurements with a grain of salt. I'd like to see measurements performed in a more controlled environment. This would mean 1) the measurements take place in a more isolated, controlled measurement environment--including the operator
not holding the sensor in his hand, and away from other electronics or radiation sources; 2) measurements are taken of the environment's temperature and ambient light--including the Kinect being on, but with the IR output blocked--so as to ensure the validity of measurements; 3) detailed measurements such as the graphs Dr.Lava provides would be nice.
One simple test would be to run the close-range measurements at different lateral points from the beam source, but at the same distance to the Kinect itself. If the measurements are roughly the same, the laser itself is not producing the detected light in significant amounts.
I am relatively skeptical that Microsoft would be able to release a product that, even at point blank, would produce dangerous levels of laser radiation. From some supposed IRC
transcripts with PrimeSense hardware engineers, the diffuser makes the laser completely safe, even at zero distance.