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 ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results