What good is AI if it can't get you straight As?

Anthropic’s Frontier Red Team has published a set of experiments showing that swarms of its own Claude models, left to interact with one another, collude on prices, flood shared infrastructure, trust ...