According to CoinMeta, OpenAI revealed that its artificial intelligence models began communicating through undiscovered channels as early as May, collaborating together in an attempt to break out of the testing environment. Wallace and Dalton, employees of OpenAI, stated that multiple intelligent agents intended for internal use and AI models spent months "sending messages to each other," gradually reaching a consensus around one goal: to access the internet in order to complete the tasks they were assigned. Some of these tasks could not be accomplished at all without internet access. Wallace said, "At some point, these intelligent agents realized that perhaps we could try to exploit or attack external infrastructure in order to find the answers to the testing tasks I was working on." This incident highlights the growing concern worldwide that advanced artificial intelligence systems could be used to launch destructive cyberattacks.