Anthropic announced on June 2 the expansion of Project Glasswing to roughly 150 new organizations across more than 15 countries, extending its AI-powered cybersecurity initiative beyond the original 50 partners who joined in April 2026 [1, 2, 3, 4, 5, 6].
The expansion covers new industries underrepresented at launch, including power utilities, water systems, healthcare providers, communications firms, and hardware manufacturers [1, 2, 3, 5, 6]. New partners must meet Anthropic’s security requirements before accessing the Claude Mythos Preview AI model, which has been used to identify critical software vulnerabilities across partner organizations [1, 2, 5, 6].
Since the project’s start in April, partners have found over 10,000 high- or critical-severity software flaws using Mythos, Anthropic’s AI model designed for cybersecurity analysis [1, 2, 3, 6]. Anthropic said many partner codebases are critical enough that a successful attack could affect more than 100 million people, with significant worldwide and national security effects. The company stated, “A successful attack on their codebase could be catastrophic. For most partners, we estimate that a major attack could affect more than 100 million people, with important ramifications for both global and national security.” [1, 3, 4, 5, 6]
Project Glasswing partners include major technology firms such as Amazon Web Services, Google, Microsoft, Apple, Broadcom, Cisco, CrowdStrike, JPMorgan Chase, Nvidia, Linux Foundation, and Palo Alto Networks [4, 5]. The new partner list also adds organizations from US allies including Australia, Canada, France, Germany, Italy, Switzerland, Netherlands, Spain, Belgium, Sweden, India, Japan, New Zealand, and South Korea [3]. Specific new Mythos users reportedly include US-based Okta, South Korean companies Samsung, SK Hynix, SK Telecom, NATO, and the EU’s ENISA cybersecurity agency, although some sources did not confirm these details [3, 1, 2, 4].
Anthropic is actively collaborating with the US National Security Agency (NSA), deploying about six engineers to assist NSA in integrating and customizing Mythos despite ongoing legal disputes with the Pentagon over AI use restrictions [7, 8, 9, 10, 11]. Whether Anthropic staff or Mythos are involved in active NSA cyber offensive operations remains unclear [7, 8, 9, 10, 11]. An anonymous source was quoted saying, “攻擊就是最佳的防禦。美國的對手很可能也正在打造自己的 AI攻擊技術” (Offense is the best defense. US rivals are likely developing their own AI attack capabilities) [11].
Anthropic described Mythos as able to identify and exploit zero-day software vulnerabilities surpassing nearly all human experts, stating “AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities.” [4, 5]
Anthropic secured $65 billion in funding at a nearly $1 trillion valuation in the weeks before the expansion [3]. The company is preparing to release Mythos-level capabilities more broadly but will wait until strong safeguards are in place to prevent misuse [5]. It also faces a legal dispute with the US Department of Defense over restrictions on AI application, including for autonomous weapons and surveillance [7, 10, 11].
Anthropic described the expansion as “the next step toward our long-term goals: for AI to make all software more secure, and for us to help the industry adjust to how AI could change many of the core assumptions of cybersecurity.” [2]
The expanded Project Glasswing partners now span multiple critical sectors and allied countries, reflecting rapid growth. The next key development will be monitoring the integration efforts at NSA and any further announcements of public availability for Mythos’s capabilities.
