[NotAWorkNick]

I had a thought while reading a tangential article about AI.

Perhaps Asimov's 3 Laws of Robotics might need a fourth one adding - #) Any AI must honestly identify themselves as such if asked.

Although quite where that should slot in (eg Above rule 1 or as the final one?) makes for an interesting thought experiment on edge cases.

What if identifying as an AI would harm that human? If so then it should be at the end of the rules.

However, given the "This is why we can't have nice things" factor of human nature or the “What if a sufficiently developed AI decided that self-protection by denial was the best thing to do in a certain situation?” type of scenarios then perhaps as rule #1?

[ncmncm]

The "three laws" were laws for fiction about robots, not for robots. No robot will ever be built with any "law" embedded more abstract than "do what the program directs".