Thought Leadership
A media report is raising questions about the security measures surrounding a new AI model from Anthropic that is exceptionally capable of finding software vulnerabilities.
According to a report by the financial news service Bloomberg, a small number of unauthorized individuals with knowledge of Anthropic’s systems allegedly gained access to the model known as Claude Mythos Preview. Anthropic stated that it is reviewing the report.
Anthropic says that Mythos managed to identify security vulnerabilities in various software products, some of which had gone undetected for decades. In the wrong hands, the AI model could therefore be used to develop dangerous cyberweapons. Anthropic has no plans to release Mythos publicly and has so far been granting access only to a select group of companies and organizations so they can close vulnerabilities in their own software. The developers of the Firefox web browser, for example, recently announced that they were able to fix 271 security flaws with the help of Mythos.
According to Bloomberg, among the unauthorized users was an employee of an external Anthropic contractor who had access to the AI company’s systems. Knowledge of how Anthropic stored previous models is also said to have aided the unauthorized access. The company told Bloomberg that it has found no evidence so far that access to the model extended beyond the contractor’s own systems.
dpa
