1 hour ago
4
Anthropic is investigating a imaginable breach of Mythos, a caller exemplary the artificial quality company rolled out to a tiny excavation of companies earlier this period to assistance them observe bundle vulnerabilities.
The AI institution down the chatbot Claude is looking into a study of unauthorized entree to Mythos from 1 of its third-party vendor environments, an Anthropic spokesperson told CBS News successful an email.
Anthropic works with a tiny fig of third-party vendors to make its AI models. So far, the institution has not detected immoderate breaches extracurricular of its vendor situation oregon immoderate compromises to the Anthropic systems.
Anthropic confirmed its probe into the imaginable Mythos breach connected Wednesday, a time after Bloomberg reported that a tiny radical of unauthorized users had gained entree to the tool, citing a idiosyncratic acquainted with the matter.
Anthropic released Mythos to a constricted radical successful April arsenic portion of an effort called Project Glasswing, billing the caller exemplary arsenic much effectual than competing AI systems astatine detecting bundle vulnerabilities.
At the time, Anthropic lone shared the instrumentality with a tiny radical of large companies, including Amazon, Apple, Cisco, JPMorgan Chase and Nvidia, amid concerns that the caller exemplary could beryllium exploited by hackers. The extremity was to assistance these companies harden their defenses earlier atrocious actors tin summation entree to Mythos oregon akin AI models.
Federal officials, information experts and leaders astatine planetary institutions similar the International Monetary Fund person each raised concerns astir what mightiness hap if Mythos falls into the incorrect hands. While Project Glasswing is intended to assistance companies insulate themselves from cybersecurity threats, immoderate experts are acrophobic that Mythos could besides beryllium utilized to exploit IT infrastructure astatine banks, hospitals, authorities systems and different organizations.
"We request to hole ourselves, due to the fact that we couldn't support up with the atrocious guys erstwhile it was humans hacking into our networks," Alissa Valentina Knight, CEO of cybersecurity AI institution Assail, antecedently told CBS News." We surely can't support up present if they're utilizing AI due to the fact that it's truthful overmuch devastatingly faster and much capable.
Edited by Aimee Picchi
In:
Key takeaways connected Anthropic's caller AI exemplary
Key takeaways connected Anthropic's concerning caller Mythos AI model
(05:06)
English (US) ·