Unauthorized Users Access Mythos Model, Anthropic Claims Technology is Highly Powerful

On April 22, according to a source familiar with the matter and documents reviewed by the media, a small number of unauthorized users have gained access to Anthropic PBC’s next-generation Mythos AI model. Anthropic has stated that the technology of this model is extremely powerful and could even be used to launch dangerous cyberattacks. Individuals who requested anonymity due to fears of retaliation indicated that on the same day Anthropic first announced plans to open testing of the Mythos model to a select few companies, a small group of users in a private online forum successfully accessed Mythos. The source noted that this group has since been regularly using Mythos but has not utilized it for cybersecurity purposes. The source also provided media with screenshots and live demonstrations of the model to corroborate these claims. This unauthorized access had not been previously disclosed, highlighting the challenges Anthropic faces in completely preventing the spread of its most powerful and potentially dangerous technology beyond authorized partners. The incident has also raised further questions about whether others are using Mythos without permission and what their intentions might be. (Zhitong Finance)