Anthropic Expands Glasswing, Promises Mythos-class Public Release

Per Anthropic's June 2 blog post, Project Glasswing is expanding by approximately 150 new organizations, bringing the partner roster to roughly 200 across more than 15 countries spanning power, water, healthcare, communications, and hardware. Anthropic said partners have so far found more than 10,000 high- or critical-severity security flaws using Claude Mythos Preview. CNBC and TechCrunch report the newly named participants include Okta, Samsung, SK Hynix, SK Telecom, NATO, and the EU cybersecurity agency ENISA, alongside earlier partners such as Apple, Nvidia, Microsoft, CrowdStrike, and Palo Alto Networks. Separately, Reuters reported on May 28 that Anthropic aims to make Mythos-class models available to customers in the coming weeks while launching Claude Opus 4.8 at unchanged pricing. Anthropic's blog cautions that broad Mythos release still depends on misuse safeguards it says have yet to be developed.
What happened
Per Anthropic's June 2 blog post, Project Glasswing, the company's restricted cybersecurity program, is expanding by approximately 150 new organizations, bringing the roster to around 200 partners based in more than 15 countries. Anthropic said the new cohort covers sectors underrepresented at launch, including power, water, healthcare, communications, and hardware, and that many partners are vendors whose codebases other organizations and governments depend on. Anthropic added that partners have used Claude Mythos Preview to find more than 10,000 high- or critical-severity security flaws since the program began in early April with roughly 50 initial partners.
Who is involved
CNBC and TechCrunch report the broader group now includes Okta, Samsung, SK Hynix, SK Telecom, NATO, and the EU cybersecurity agency ENISA, alongside earlier partners such as Apple, Nvidia, Microsoft, CrowdStrike, and Palo Alto Networks. Each organization must meet Anthropic's security requirements before gaining access.
Opus 4.8 and the rollout question
Reuters reported on May 28 that Anthropic aims to make Mythos-class models available to customers in the coming weeks, issued alongside the launch of Claude Opus 4.8. Per Reuters and TheNextWeb, Opus 4.8 ships at the same price as its predecessor and posts benchmark gains, with Anthropic emphasizing improved honesty - the model is described as more likely to flag uncertainty and less likely to make unsupported claims. Anthropic's own blog is more guarded on timing, stating that safely releasing Mythos-level capabilities in general access will require misuse safeguards that it and, to its knowledge, other developers have yet to develop.
Why it matters
The reported discovery of more than 10,000 serious flaws in a controlled cohort is a concrete data point on how quickly frontier models can surface vulnerabilities across large codebases. For defenders, wider Mythos access could accelerate scanning and patching; the same capabilities, released without safeguards, could lower the barrier for attackers. That dual-use tension is now driving public debate over release controls, disclosure timelines, and coordination with governments.
What to watch
- •The timing and access conditions for any broader Mythos availability beyond Glasswing.
- •Whether partners publish remediation or coordinated-disclosure data following scans.
- •How national-security stakeholders respond, given partners now include NATO and ENISA.
Key Points
- 1Project Glasswing is expanding to about 150 new partners, roughly 200 in total across 15-plus countries, widening defensive access to Mythos-class capabilities for critical-infrastructure providers.
- 2Anthropic reports partners have surfaced more than 10,000 high- or critical-severity flaws, a material signal of how fast frontier models can find vulnerabilities at scale.
- 3Anthropic frames broad Mythos release as contingent on misuse safeguards it says do not yet exist, underscoring the dual-use tension as high-capability cyber models approach general availability.
Scoring Rationale
A controlled-program expansion of a high-capability cybersecurity model to roughly 200 partners, paired with reporting on a broader Mythos rollout and the Opus 4.8 launch, is materially important to security-focused ML practitioners and infrastructure teams. It is a significant industry development rather than a paradigm-shifting model launch, supporting a score in the low-to-mid 7s.
Sources
Public references used for this report.
View 7 more sources
- 04Anthropic scales Claude Mythos to critical infrastructure in 15+ countriestechcrunch.com
- 05Anthropic’s Guarded Mythos Model Is Headed For Wider Releaseforbes.com
- 06Anthropic’s Claude Opus 4.8 is its most honest AI model yet, and Mythos is coming in weeksthenextweb.com
- 07Anthropic expands Glasswing as it promises public Claude Mythos-class model releases9to5mac.com
- 08Anthropic is bringing Claude Mythos to India, select organisations to get preview accessindiatoday.in
- 09Project Glasswing Grows as Anthropic Extends Claude Mythos Preview Programitsecuritynews.info
- 10India gets access To Claude Mythos AI model as Anthropic expands the initiativethehindubusinessline.com
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems

