Anthropic invited approximately 150 more organizations to participate in Project Glasswing, which provides access to its new model, Claude Mythos. Previously, the model was available only to select organizations since its announcement in early April.
Project Glasswing aims to leverage the capabilities of Claude Mythos, with Anthropic stating that AI models have reached a level where they can detect and exploit software vulnerabilities at a higher level than most skilled human engineers. The initial announcement saw nearly 50 partners, including major companies like Apple, Amazon, Broadcom, CrowdStrike, Microsoft, and NVIDIA.
The newly invited organizations come from over 15 countries and include sectors such as public utilities and healthcare, which were not well represented in the initial group. However, they must meet Anthropic’s security requirements before gaining access to Claude Mythos.
Anthropic highlighted the potential catastrophic impact of successful cyber-attacks on its partners’ codebases, estimating that a major attack could affect over 100 million people, with significant implications for both global and national security.
The company indicated it is working to release Mythos capabilities to the public but stated that it would implement “highly robust safeguards” to prevent misuse beforehand. In May, the announcement of Claude Mythos influenced a reconsideration of AI regulation by the Trump administration, which was anticipated to sign an executive order aimed at facilitating the sharing of AI-identified software vulnerabilities with critical institutions. However, the signing was postponed due to objections from Trump and the tech industry.
