US-based AI developer Anthropic expanded its cybersecurity initiative, Project Glasswing, on Tuesday to protect critical infrastructure across more than 15 countries using its powerful Claude Mythos model.
This expansion follows the company’s confidential filing for an initial public offering (IPO) on Monday, hot on the heels of a $65 billion funding round that valued the artificial intelligence startup at nearly $1 trillion.
At the core of Project Glasswing is Claude Mythos, which Anthropic describes as its most capable model to date. The system can scan vast codebases to detect thousands of zero-day software vulnerabilities within weeks. In early April, Anthropic initiated the project by granting 50 select partners, including the United States government, access to the Claude Mythos Preview to identify security flaws.
Securing Critical Infrastructure Worldwide
The latest expansion brings approximately 150 new organisations into the fold. This cohort focuses heavily on sectors previously underrepresented in the trial, such as power grids, water networks, healthcare systems, telecommunications, and hardware manufacturing. Anthropic highlighted that many of these new participants are companies or non-profit organisations maintaining foundational codebases that other enterprises and governments depend upon daily.
The stakes are incredibly high. Anthropic warned that a successful breach of these codebases could have devastating consequences, estimating that a major attack on just one partner could impact over 100 million people and severely compromise both national and international security.
Global Reach and Key Partners
The newly added organisations span several allied nations, including the United Kingdom, Australia, Canada, France, Germany, Italy, Switzerland, the Netherlands, Spain, Belgium, Sweden, India, Japan, New Zealand, and South Korea, according to a report by the Financial Times citing a source close to the matter.
Among the high-profile entities granted access to the Mythos model are the US identity management platform Okta, South Korean tech giants Samsung, SK Hynix, and SK Telecom, the Brussels-headquartered NATO military alliance, and ENISA, the European Union’s cybersecurity agency. TechCrunch has contacted Anthropic to confirm these details.
The Race Against AI-Powered Cyber Threats
Anthropic is moving quickly because it anticipates that competing AI developers will soon produce models with capabilities matching Mythos Preview. Project Glasswing is designed to establish vital defensive safeguards before these advanced capabilities become widespread.
The competitive pressure is mounting. Since the debut of Mythos, rival firm OpenAI has launched its own cybersecurity-oriented model, GPT-5.5-Cyber, which is currently being trialled by a select group of partners.