Anthropic’s Mythos AI breach: Users report unauthorized access

For past weeks, various tech experts and financial chiefs have raised concerns over potential misuse of Anthropic’s new powerful AI model: Claude Mythos Preview.

Now their fears have proven justified when Bloomberg reported on Wednesday that Mythos AI model, equipped with unprecedented cyber abilities, has been accessed by a small group of unauthorized users.

According to a source privy to the matter, a handful of users belonging to a private online forum gained access to Mythos when Anthropic decided to release the model to few tech companies under Project Glasswing.

More worryingly, since then the unauthorized users have been using Mythos for various purposes.

"We are investigating a report claiming unauthorized access to ‌Claude ⁠Mythos Preview through one of our third-party vendor environments," an Anthropic spokesperson said.

Since its release, Anthropic has touted Mythos AI as a powerful model which can identify and exploit critical vulnerabilities in every “major operating system and web browser when directed by a user to do so.” Given its capabilities, the US-based artificial intelligence company has decided to curtail its public-wide release.

Anthropic CEO warned, if misused, Mythos can become a serious threat to economies, public safety and national security.

In recent days, the major financial institutions and chiefs of the banking system have voiced concerns over risks posed by the AI model. They are also urging to be included in the list of early testers in order to protect their own systems.

Moreover, the recent report of unauthorized access highlights security loopholes and challenges Anthropic faces in fully preventing the access of its most powerful model beyond approved partners.