Australia Gets "Dangerous" AI: Claude Mythos Preview Access

Australian organizations are gaining access to an AI model called Claude Mythos Preview, which its creator, Anthropic, previously deemed "too dangerous" for public release. This is part of a global rollout, known as Project Glasswing, aimed at strengthening critical systems against hackers. Anthropic expanded access to about 150 new organizations across more than 15 countries. Mythos is highly skilled at identifying software vulnerabilities. Anthropic has withheld it from general release due to concerns its cyber capabilities could be used to attack financial systems and infrastructure. Australian organizations will receive access, though specific names were not disclosed. Anthropic has previously briefed institutions like the Treasury and the Reserve Bank. This development comes as competitors are also working on powerful AI models. This matters because it highlights the ongoing tension between powerful AI capabilities and potential security risks.