Smaller-scale AI deception incidents (like current LLM lying detected in research) serve as valuable scientific evidence and opportunity for safety fixes, even though they are unlikely to trigger government regulation because they do not cause visible human harm.