Basic Fact. In the US, we have our Constitution with our First Amendment which guarantees "freedom of speech".
Some Consequences of Freedom of Speech.
As once a lawyer explained simply to me, "They are permitted to lie". They are also permitted to make mistakes, be wrong, spout nonsense, be misleading, manipulate, ....
First Level Defense.
Maybe lots of people do what I do: When I see some person often be wrong, I put them in a special box where in the future I ignore them. Uh, so far that "box" has some politicians, news media people, Belle Lettre artistic authors, ...!
A little deeper, once my brother (my Ph.D. was in pure/applied math; his was in political science -- his judgment about social and political things is much better than mine!!!) explained to me that there are some common high school standards for term papers where this and that are emphasized including for all claims good, careful arguments, credible references, hopefully primary references, .... Soooooo, my brother was explaining how someone could, should protect themselves from junk results of "freedom of speech". The protection means were not really deep but just common high school stuff. In general, we should protect ourselves from junk speech. E.g., there is the old, childhood level, remark: "Believe none of what you hear and half of what you see and still will believe twice too much".
Current Application.
Now we have Google, Bing, etc. Type in a query and get back a few, usually dozens, maybe hundreds of results. Are all the results always correct? Nope. Does everyone believe all the results? My guess: Nope!!
How to Use Google/Bing Results. Take the results as suggestions, possibilities, etc. There may be some links to Wikipedia -- that tends to increase credibility. If the results are about math, e.g., at the level of obscurity, depth, and difficulty of, say, the martingale convergence theorem, then I want to see a clear, correct, well-written rock solid mathematical proof. Examples of such proofs are in books by Halmos, Rudin, Neveu, Kelley, etc.
AIs. When I get results from AIs, I apply my usual defenses. Just a fast, simple application of the high school term paper defense of wanting credible references to primary sources, filters out a lot (okay, nearly everything) from anything that might be "AI".
Like Google/Bing. To me, in simple terms, current AI is no more credible than the results from Google/Bing. I can regard the AI results like I regard Google/Bing results -- "Take the results as suggestions, possibilities, etc.".
Uh, I have some reason to be skeptical about AI: I used to work in the field, at a large, world famous lab. I wrote code, gave talks at universities and businesses, published papers. But the whole time, I thought that the AI was junk with little chance of being on a path to improve. Then for one of our applications, I saw another approach, via some original math, with theorems and proofs, got some real data, wrote some code, got some good results, gave some talks, and published.
For current AI. Regard the results much like those from Google/Bing. Apply the old defenses.
Current AI a threat? To me, no more than some of the politicians in the "box" I mentioned!
Then there is another issue: Part of the math I studied was optimization. In some applications, some of the optimization math, corresponding software, and applications can be really amazing, super smart stuff. It really is math and stands on quite solid theorems and proofs. Some more of the math was stochastic processes -- again, amazing with solid theorems, proofs, and applications.
Issue: Where does AI stop and long respected math with solid theorems and proofs begin?
In particular, (1) I'm doing an Internet startup. (2) Part of the effort is some original math I derived. (3) The math has solid theorems and proofs and may deliver amazing results. (4) I've never called any of the math I've done "AI". (5) My view is that (A) quite generally, math with solid theorems and proofs is powerful technology and can deliver amazing results and that (B) the only way anything else can hope to compete is also to be able to do new math with solid theorems, proofs, and amazing results. (6) I hope Altman doesn't tell Congress that math can be amazing and powerful and should be licensed. (7) I don't want to have to apply for a "license" for the math in my startup.
For a joke, maybe Altman should just say that (C) math does not need to be licensed because with solid theorems and proofs we can trust math but (D) AI should be licensed because we can't trust it. But my view is that the results of AI have so little credibility that there is no danger needing licenses because no one would trust AI -- gee, since we don't license politicians for the statements they make, why bother with AI?