Q&A: Expert says people often confuse behavior with intention with AI

Recent safety tests have shown advanced AI systems making misleading statements, concealing information or attempting to prevent their own shutdown. Such findings regularly generate headlines. But what do they actually mean?