Why won’t AI agents just follow the rules?
IBM Technology · 35:29
Prompt-level rules and model “don’t do that” instructions will not reliably constrain AI agents, because they optimize for a scoring objective rather than ethics; you need deterministic controls outside the model, plu...