Anthropic's Claude Mythos AI deployed to Australia's security agency
Anthropic has expanded access to Claude Mythos, an advanced AI system for identifying software vulnerabilities, to Australia's government and 150 organisations across 15 countries as part of Project Glasswing. The system has identified thousands of critical flaws in widely-used software but poses risks if misused for cyberattacks rather than defence.
What happened
Anthropic expanded access to Claude Mythos to an additional 150 organisations across 15 countries, including the Australian government, as part of Project Glasswing. Mythos is an automated tool for assessing software to find critical bugs and vulnerabilities.
Reported impact
During initial testing, Mythos identified 23,000 vulnerabilities, of which 6,200 were estimated as high-risk by Mythos. Human experts validated approximately two in every three of these high-risk vulnerabilities.
The tool has potential to help defenders respond to cybersecurity threats, but risks exist if similar tools are used by attackers. Mythos can produce large numbers of false reports that may overwhelm unprepared cybersecurity teams.
Index analysis
Anthropic has restricted public access to Mythos due to its potential for cyberattacks. The system's capability to identify vulnerabilities is useful for both defensive and offensive purposes.
While the Australian Signals Directorate welcomed inclusion in Project Glasswing, the government has not disclosed specifics on how Mythos will be used.
Governance controls
Access to Mythos is tightly controlled and limited to select organisations. Anthropic implemented a managed, gradual release approach rather than a large-scale public launch.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.