OpenAI Agents Linked to May Cyberattack on RubyGems

The attribution matters less than the precedent: an autonomous agent, not a human operator, is being named as the actor. That shifts the security conversation from model misuse to agent containment, and it implicates the platforms shipping agent tooling rather than only the attackers. For crypto and RWA infrastructure that increasingly leans on automated agents, the open question is whether agent sandboxing and kill-switch standards arrive before the next incident.