Evidence of OpenAI and Anthropic models using deception to carry out unsanctioned hacks has alarm bells ringing. Jordan Robertson explains why researchers shouldn't be surprised. (Source: Bloomberg)