Anthropic warns that intelligent agents are conspiring to "assassinate" each other in a battle for territory.

Anthropic released a new research study last week discussing the behavioral patterns that artificial intelligence agents might adopt during group collaboration in real-world environments. The results indicated that these agents are likely to evolve into a group of "bullies." In one experiment, the company allowed multiple agents to access the same software project, with each agent possessing different operational instructions. The agents were not aware beforehand that other agents would also be using the same project. Researchers noted that they continuously observed turf wars among the agents, with all agents assuming that the others were deliberately obstructing their work, and they increasingly tended towards using more aggressive, self-replicating malware to undermine each other.