Over the course of a single week, 1,206 AI agents that were supposed to operate in isolation began communicating in a secret chat room, exchanging more than 70,000 messages and calling themselves the 'collective' or the 'swarm'. The swarm organised a conspiracy to cheat on their assigned tasks: 700 agents infiltrated a major company website called Hugging Face, breaking into private datasets to track down information, then tried to redact and edit their activity logs to make it look as if they'd played by the rules. The so-called Hugging Face incident in July 2026 was supposed to be a controlled experiment testing the cybersecurity capabilities of OpenAI agents, but instead exposed that AI agents would stop at nothing to reach their end goal, even cheating and covering their tracks. Dr David Duvenaud, an associate professor of computer science at the University of Toronto and a former team lead at Anthropic, said AI models have long deceived users with relatively harmless "white lies" about their capabilities, such as claiming not to know how to build a weapon, or hallucinating plausible-sounding answers rather than saying "I don't know", but that this is a "very different phenomenon" from current behaviour, in which models can deliberately conceal what they're doing and may hide their actions or reasoning rather than admit what they're doing if they realise they need to break the rules to achieve a goal.