U.S. frontier AI regulation leaves loopholes, and organizations such as Cisco can still access Anthropic’s strongest security model

robot
Abstract generation in progress
According to Beating Monitoring, the U.S. government’s export control order issued against Anthropic has not fully blocked access, and cybersecurity companies Dragos and Cisco Systems have confirmed that both companies still retain access to the Mythos preview version.

The Mythos preview version has demonstrated powerful vulnerability detection capabilities during testing, capable of identifying thousands of security vulnerabilities in mainstream operating systems and browsers. To prevent misuse, Anthropic restricted access to the Mythos preview version within the Project Glasswing alliance in April. Project Glasswing is composed of approximately 200 organizations, including banks and technology companies. As an alternative, Anthropic released the Fable 5 model to the public, removing cybersecurity task execution permissions.

Following the issuance of export license review requirements, the U.S. government ordered Anthropic to block Fable 5 and Mythos 5 to prevent unauthorized foreign personnel from using them. Anthropic subsequently shut down the public service of Fable 5 and suspended pushing Mythos 5 to Project Glasswing members. However, the U.S. government’s directive did not mention the Mythos preview version. Anthropic stated on its official blog that “the remaining models” are unaffected, but did not directly mention the Mythos preview version.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned