Trump Orders Faster Access to Cyber AI Models for Partners While Revoking Public Release

President Trump signed an executive order this month directing faster access to advanced cyber AI models for partner organizations and government agencies. Anthropic released Claude Mythos Preview in April 2026. 5-Cyber shortly afterward.

Both companies limited the unrestricted versions to a small number of partner organizations and government agencies. The Trump administration later required Anthropic to revoke all public access to the latest version of Mythos. Palo Alto Networks recorded a fourfold increase in daily cyberattacks among its clients between 2024 and 2025.

Moody’s Ratings reported that the time for attackers to exploit a known vulnerability dropped from more than 700 days in 2020 to 44 days in 2025. Anthropic used Claude Mythos Preview to identify thousands of bugs in open-source software packages. Mozilla applied the same model to correct more than 400 bugs in the Firefox browser during April 2026, compared with its usual rate of roughly 20 fixes per month.

The courseware platform Canvas was breached last month, disrupting classrooms in thousands of schools and universities. The group ShinyHunters, which carried out the Canvas attack, later compromised an Oracle HR system and may have taken data from more than 100 organizations.

A group of cybercriminals used Meta’s customer-service AI to obtain access to approximately 30,000 Instagram accounts, including the Sephora corporate account and the former Obama White House account.

Meta spokesperson Andy Stone said internal backend checks failed but the incident was not caused by the AI agent itself and the underlying issue has been addressed. Amazon recorded multiple outages in its e-commerce services that were attributed to AI-generated code. Anthropic maintains about 200 partner organizations that use its cyber AI models.