Unauthorized Users Access Anthropic's Powerful AI Tool, Mythos

Anthropic's Mythos, a cybersecurity tool that can identify vulnerabilities in major operating systems, has been accessed by an unauthorized group. The group gained access through a third-party contractor portal and internet sleuthing tools, according to Bloomberg [doc2]. Anthropic is investigating the incident, but no evidence of system impact has been found. CEO Sam Altman criticized Anthropic's fear-based marketing, suggesting it may be trying to sell the tool to select customers without releasing it publicly [doc13].

A group of unauthorized users has gained access to Anthropic's powerful AI tool, Mythos, which is designed to identify vulnerabilities in major operating systems and browsers . The incident came to light when Bloomberg reported that the group, consisting of individuals with access to a contractor portal, used common internet sleuthing tools to breach the system .

Anthropic has acknowledged the incident and stated that it is investigating the claims, but has not found any evidence to suggest its systems have been impacted .

This incident has raised concerns within the tech community, as Mythos has been touted for its potentially significant capabilities in cybersecurity and security concerns may spread beyond China.

Notably, Anthropic has chosen not to publicly release Mythos due to its potential risks, which has sparked criticism from OpenAI CEO Sam Altman. He accused the company of using 'fear-based marketing' by selling exclusive access to the tool to select customers, likening it to selling a $100 million bomb shelter while threatening to drop a nuclear bomb .