By Andrew Cullen
Publication Date: 2026-06-04 23:35:00
Artificial intelligence (AI) giant Anthropic has expanded access to a sophisticated model deemed too dangerous to release publicly, including Australia among its select handful of users.
The large language model, known as Claude Mythos, is now being rolled out to an additional 150 organizations in 15 countries as part of Project Glasswing, including the Australian government and several local companies.
At a time when large-scale AI deployments are occurring daily, this limited, phased release may seem particularly surprising. But Mythos is not like most other AI systems. Instead, it is an automated tool for evaluating software to find critical bugs and vulnerabilities.
This managed release is intentional because the discovery of vulnerabilities in computer systems is useful both to those who want to defend them and to those who want to hack them.
However, the actual impact of AI systems on cybersecurity is much more complex.
