A small group of unauthorized users has already accessed and is regularly using Anthropic's restricted Mythos Preview AI model, though not for cybersecurity purposes as the model is intended, Bloomberg reported on April 21.

Anthropic has claimed it is investigating Bloomberg's reporting but preliminarily stated there is no evidence that the access extended to Anthropic's systems. Anthropic preliminarily released Mythos on April 7 to major companies like AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, Nvidia and Palo Alto Networks. According to Anthropic, Mythos can discover vulnerabilities in every major operating system and web browser. The "handful" of unknown users reportedly gained access to the model the same day Anthropic made its limited release to companies for testing purposes.