"Prompt injection" is a very good name, because it's ultimately the same problem as SQL injection or cross-site scripting. In SQL injection, you have two layers of meaning: The SQL query itself, and the data contained in the query. The injection is when you incorrectly allow a user's input to cross the boundary between the two, and the data becomes structural SQL query.
In this case, in order to make an "ethical AI", what they need to do by their own definition is modify the underlying neural net to be unable to emit anything "bad". Unfortunately, this is fundamentally impossible since the neural nets are opaque. So it looks like these systems try to work by feeding the AI a prompt behind the scenes telling it all about how it won't be naughty. But that's the wrong layer. It's the same layer user input will be on. The fact that the right thing to do to solve this problem is impossible is not a concern of the algorithm or implementation. It just means the right thing can't be done.
This basically can't work, and honestly, this is going to be a real problem. "Public" AI research is going to constantly be hogtied by the fact that if the AI does something bad, we blame the AI and not the user trying to trick it. I assure you, private AI research is proceeding without any such constraints or problems.
It is too much to expect a 2022 AI to 100% correctly filter out things that violate Silicon Valley Liberal dogma, or any other political dogma. That is not a thing this technology is capable of. That's a superhuman problem anyhow. It is mathematically not possible with the current technologies; the intrinsic biases of the systems are not capable of representing these sensibilities. So, either start putting the word around that people who trick the AI into saying crazy things are themselves the source of the crazy and you should stop blaming the AI... or stop putting the AIs on the internet. Because there is no third option. There is no option where you can put a safe, sanitized AI that can't be tricked into doing anything X-ist. The technology isn't good enough for that. It wouldn't matter if you scaled them up by a hundred times.