Ethics of AI – for humans or for AI?

Recently, more and more media coverage has appeared on how clients and public institutions react to the “products” of professional advisers and legal representatives created with or by AI. And the reactions have been, to put it mildly, far from positive.

Starting from a recent case of a consulting firm from the Antipodes and ending with a newly published decision of the Polish National Appeal Chamber (KIO) excluding a bidder whose explanations regarding abnormally low pricing were drafted by AI – the reactions have been uniformly negative.

In both cases, the issue was not the mere fact that the document had been prepared by AI, but that it contained references to non-existent case law or scholarly opinions.

Anyone who has worked with AI even briefly knows that it is not a tool capable of replacing human work. At best – even when “fed” with the most sophisticated prompts – AI should be treated as a junior trainee.

So, is the problem the use of AI itself, or rather the misuse of the tool?

Here lies the grey area: if someone uses AI responsibly, reviews its output, and adapts it using their experience and expertise, can anyone really determine how much of the final result was AI’s work? In practice, this is nearly impossible.

The real “AI problems” are still ahead of us, yet doubts are already arising:

• Should AI-co-created material include a disclaimer?
• Should consulting firms openly inform clients about using AI in their work?
• Should the fees for AI-assisted services be the same as for human work?

As with every new phenomenon, time will tell – and moderation will remain key.

Isaac Asimov once formulated three laws of robotics (later complemented by a fourth, equally reasonable one). Could similar rules be “encoded” into AI itself?

Because the ethical principles of AI creation followed by developers seem, given AI’s growing autonomy, far from sufficient.

Leaving these questions open, I’ll share one concern: AI is already so widespread that I can’t rule out the possibility that we have reached a point where, during correspondence between parties – whether in a construction project or even a court proceeding – AI is already sitting on both sides of the table, while we humans merely observe and (more or less) control what is happening.

Hopefully, more rather than less.