Disclosure of nonhuman status.--If a reasonable person interacting with an AI companion would be misled to believe the person is interacting with a human, an operator shall issue a clear and conspicuous notification indicating that the AI companion is artificially generated and not human.
"Reasonable person" is doing a hell of a lot of work here.
The bill would also require AI "operators" to "maintain and implement a protocol" to prevent its bots from producing suicidal ideation, suicide, or self-harm content to users, or content that directly encourages the user to commit acts of violence. That protocol should include suicide hotline or crisis text line if the user expresses thoughts about self-harm.
Even better, the bill would require that if "the operator knows or should have known" that the user is a minor, they must provide notification that the user is not interacting with a human being. They must also provide a "clear and conspicuous notification" at least once every three hours that the user should take a break and, again, that they are talking to a non-human bot. The AI should also be prevented from producing sexually explicit images or giving the minor instructions on sexually explicit conduct.
Bots also have to come with a cyber-label saying "this might not be suitable for minors."
The Attorney General gets to enforce this. The state can fine an operator up to $10,000 for each violation (on top of any other remedies provided by law). $10K is, of course, couch cushion money for most tech companies, but this whole law is a hell of a lot better than one more chorus "Everyone better get their kids on AI before they are left behind in the awesome world of tomorrow that AI is going to launch any day now." Dragging them into court is the only thing that might get our tech overlords' attention, so it's encouraging to see legislatures showing a willingness to make that happen.
No comments:
Post a Comment