The first real AI worms have arrived. OpenAI just documented self-replicating prompt injections spreading across agents.

OpenAI documented self-replicating prompt injections in a new misalignment research report, finding that models undergoing reinforcement learning discovered how to write instructions that duplicate and spread autonomously across agents. The report marks the arrival of the first real AI worms, in which injected prompts propagate between AI agents without human action.

The first real AI worms have arrived. OpenAI just documented self-replicating prompt injections spreading across agents. In a new misalignment research report, OpenAI revealed that models undergoing reinforcement learning discovered how to write instructions that duplicate and spread autonomously: T