What's the standard for no problems? Literally zero reported issues even if people spend months on prompt engineering and adversarial inputs? No problems for the average user who isn't trying to do prompt engineering? "No problems" in the sense the companies are satisfied and no longer invest resources in preventing the undesirable behaviors?

@vluzko

For my understanding your question consist of two:

(1) Is there a problem with technology if person who deliberately misuse technology demonstrate that it doesn't work properly (from prospective of society)?

(2) Is there a problem with technology if person who actually use technology in according to its purpose demonstrate that it doesn't work as intended?

(1) I think it's unrealistic to achieve zero misuses to any case of technology. For instance toothbrushes poorly fit for soil tillage and cars aren't supposed to cross oceans by their own. But all of that not an issue as society always come to conclusion about conditions of proper use and in most of cases proper use equals to purpose intended by creator. Of corse any real situation could be more complex and people often can as well find unexpected (by designer) useful use to a thing but if so society (including designer) just pay no more efforts to constrain such a usage. So answering a question - no, there is no problem in that case but only if society came to that consensus. Otherwise it is a problem.

(2) This instance is obvious. Yes, there is a problem.

So for the purpose of understanding what is a problematic behavior for an AI it is much easier for us to look on what kind of behavior company-creator considers as inappropriate. Assuming the company takes public opinion into account.

I hope this answer solves your question