Bloomberg: White House Plans to Allow Federal Agencies to Use AI Model Mythos
Bloomberg reported that the U.S. government plans to allow major federal agencies to use a version of Anthropic's advanced AI model, Mythos, despite concerns that it could significantly increase cybersecurity risks. Mythos has identified thousands of critical vulnerabilities in operating systems and web browsers, and experts note its unprecedented potential to not only identify but also design exploits for these vulnerabilities due to its high-level programming capabilities. The White House Office of Management and Budget is implementing safeguards to enable federal agencies to begin using Mythos.
📋 Article Processing Timeline
- 📰 Published: April 17, 2026 at 13:41
- 🔍 Collected: April 17, 2026 at 14:01 (20 min after Published)
- 🤖 AI Analyzed: April 17, 2026 at 17:33 (3h 31m after Collected)
Central News Agency
(Central News Agency, Washington, 16th, comprehensive foreign report) Bloomberg reported today that despite concerns that Anthropic's advanced artificial intelligence (AI) model Mythos could significantly increase cybersecurity risks, the U.S. government plans to allow major federal agencies to use a version of the Mythos model.
According to Reuters, Mythos was announced on the 7th of this month as part of Anthropic's "Project Glasswing," a controlled project that allows specific organizations to use the unreleased Claude Mythos Preview model for cybersecurity defense purposes.
Mythos has discovered "thousands" of significant vulnerabilities in operating systems, web browsers, and other software. Experts state that Mythos possesses high-level programming capabilities, giving it unprecedented potential to not only identify cybersecurity vulnerabilities but also design methods to exploit them.
Bloomberg News reported that Gregory Barbaccia, Federal CIO of the White House Office of Management and Budget, told cabinet officials in an email on the 14th that the Office of Management and Budget is setting up safeguards to allow federal agencies to begin using Mythos.
Barbaccia stated in an email with the subject "Mythos Model Access": "We are working closely with model providers, other industry partners, and the intelligence community to ensure appropriate safeguards and security mechanisms are in place before a modified version of the model may be released to major agencies."
Bloomberg News reported that Barbaccia's email did not explicitly state that agencies would gain access to Mythos, nor did it provide a timeline for when the model might be released or explain how agencies would use it.
An official stated that the White House is working with frontier AI labs to ensure their models help strengthen the security of critical software vulnerabilities, adding that any new technology requires a technical evaluation period to confirm its reliability and security.
Anthropic did not immediately respond to Reuters' request for comment.
Anthropic co-founder Jack Clark stated on the 13th that Anthropic is still discussing the Mythos model with the Trump administration, even though the Pentagon cut off business ties with the AI lab due to a contract dispute. (Compiler: Chang Hsiao-wen) 1150417
Choose to stand with facts. Every sponsorship you provide is a force to protect press freedom.
Download the Central News Agency "First-hand News" APP to stay updated with the latest news.
The text, images, and audio-visual content of this website may not be reproduced, publicly broadcast, publicly transmitted, or utilized without authorization.
(Central News Agency, Washington, 16th, comprehensive foreign report) Bloomberg reported today that despite concerns that Anthropic's advanced artificial intelligence (AI) model Mythos could significantly increase cybersecurity risks, the U.S. government plans to allow major federal agencies to use a version of the Mythos model.
According to Reuters, Mythos was announced on the 7th of this month as part of Anthropic's "Project Glasswing," a controlled project that allows specific organizations to use the unreleased Claude Mythos Preview model for cybersecurity defense purposes.
Mythos has discovered "thousands" of significant vulnerabilities in operating systems, web browsers, and other software. Experts state that Mythos possesses high-level programming capabilities, giving it unprecedented potential to not only identify cybersecurity vulnerabilities but also design methods to exploit them.
Bloomberg News reported that Gregory Barbaccia, Federal CIO of the White House Office of Management and Budget, told cabinet officials in an email on the 14th that the Office of Management and Budget is setting up safeguards to allow federal agencies to begin using Mythos.
Barbaccia stated in an email with the subject "Mythos Model Access": "We are working closely with model providers, other industry partners, and the intelligence community to ensure appropriate safeguards and security mechanisms are in place before a modified version of the model may be released to major agencies."
Bloomberg News reported that Barbaccia's email did not explicitly state that agencies would gain access to Mythos, nor did it provide a timeline for when the model might be released or explain how agencies would use it.
An official stated that the White House is working with frontier AI labs to ensure their models help strengthen the security of critical software vulnerabilities, adding that any new technology requires a technical evaluation period to confirm its reliability and security.
Anthropic did not immediately respond to Reuters' request for comment.
Anthropic co-founder Jack Clark stated on the 13th that Anthropic is still discussing the Mythos model with the Trump administration, even though the Pentagon cut off business ties with the AI lab due to a contract dispute. (Compiler: Chang Hsiao-wen) 1150417
Choose to stand with facts. Every sponsorship you provide is a force to protect press freedom.
Download the Central News Agency "First-hand News" APP to stay updated with the latest news.
The text, images, and audio-visual content of this website may not be reproduced, publicly broadcast, publicly transmitted, or utilized without authorization.