AI security systems pipe video from cameras to servers which use machine learning programs to try to recognize threats. People worry that these secret, nonfree AI programs are biased.

A company that sends such data to someone else's server to be analyzed is making itself the victim of SaaSS (Service as a Software Substitute).This is equivalent, in denying that company the control of its own computing, to running a nonfree program.

The article doesn't mention the worst threat of these systems: videos transmitted to a server can be recorded by the NSA, which makes this a very dangerous massive surveillance system.

Of course, we know better than to trust the NSA to be "good guys". Don't be a sucker!

Operating a video camera that is connected to a network, or that can make the video available other than locally, should be forbidden by law except when authorized by a specific court order.

On the other hand, if the camera is locally connected to a computer that runs a free machine learning to analyze the video, and which doesn't send the video anywhere else, that seems basically legitimate as long as the program's behavior is not biased. Given that the program is free, it will be possible to check for bias and correct it.