Unauthorized users access Anthropic's Mythos AI tool

10 articles · Updated · Bloomberg · Apr 30

The breach involved a small group in a private online forum, according to a person familiar with the matter and documents reviewed by Bloomberg News.
Anthropic has restricted Mythos to carefully selected parties because the system is highly effective at identifying software and computer-system vulnerabilities.
The company warns wider misuse could help attackers steal data or disrupt critical infrastructure, intensifying security concerns around advanced AI cyber tools.

Is a leaked hacking AI our best defense or the start of a digital apocalypse?

With AI finding flaws faster than humans can fix them, is a global 'vulnerability tsunami' now inevitable?