AI ethics
The line between "lawful" and "defensible" — and what to say when someone asks why.
Compliance answers "are we allowed to". It does not answer the question that arrives right after: will you be able to explain this decision to a customer, an employee and a journalist who ask why the system did what it did.
This is not a values workshop. It is a set of concrete decisions: where the human stands in the process, what we record so a decision can be reconstructed, and who has standing to say no after the model has said yes.
A common argument
„AI ethics is a soft topic with no legal consequences.”
How it actually is
It has them indirectly and painfully: what you promised in a policy, in your terms and in marketing becomes the yardstick your conduct is measured against. A promise without a process is evidence against you.
Questions I hear most often
- ? Where does the human actually stand in our process?
- ? What must we log to reconstruct a decision six months later?
- ? How do we tell customers they are talking to a system?
- ? Who inside the company can stop a rollout, and on what basis?