![]() |
| The logo of Anthropic's artificial intelligence (AI) model "Claude Mythos" displayed on a smartphone screen. /AFP, Yonhap |
Anthropic, the developer of Claude, announced on June 30 (local time) that the US administration has lifted the export controls imposed on its highest-tier artificial intelligence (AI) models, Fable and Mythos. This comes 18 days after user access was suspended.
Anthropic stated via social media on that day that it received a notification from the Department of Commerce lifting the access restrictions on both models. For Mythos, restrictions had already been partially eased on June 26.
Anthropic explained, "'Claude Fable 5' will be available again worldwide starting tomorrow. Following a series of discussions with the US administration, we decided to redeploy the models by applying a new classifier system that can more accurately detect and block cybersecurity threat activities."
The company added, "We plan to continuously upgrade the classification system over the coming weeks to reduce false positives and better distinguish actual abuse cases from legitimate requests."
Furthermore, it noted, "Together with global AI security collaboration project 'Glasswing' partners—including Amazon, Microsoft, and Google—we have begun evaluating the severity of AI 'jailbreaking' (bypassing security safeguards) and drafting a consensus framework on how AI developers should respond."
According to Reuters, the administration had strengthened its oversight of newly released models to censor potential threats, over concerns that military intelligence agencies of countries of concern, such as China and Russia, could misuse advanced AI models.
The Department of Commerce had imposed a control regulation via a private letter on June 12, demanding an immediate restriction on foreign access to Mythos and Fable due to potential security threats. In response, Anthropic blocked access for all users as it lacked a mechanism to verify the nationality of individual users.
Kim Hyun-min
1
2
3
4
5
6
7