granting access<\/a> to GPT-5.4-Cyber, where Cloudflare, CrowdStrike, iVerify, Palo Alto Networks, SpecterOps and Zscaler are among the early security partners. OpenAI said it has clear, objective criteria and methods such as KYC and identity verification to guide who can access\u2060 more advanced capabilities.<\/p>\nUnlike Mythos, which Anthropic said it an entirely new class of AI model, GPT-5.4-Cyber is a version of OpenAI’s publicly available GPT\u20115.4 which lowers the refusal boundary for legitimate cybersecurity work and enables new capabilities for advanced defensive workflows. It’s thus unsurprising the threshold for accessing Mythos would be higher than for GPT-5.4-Cyber.<\/p>\n
The two pure-play cybersecurity vendors to serve as launch partners for Anthropic’s Project Glasswing – CrowdStrike and Palo Alto Networks – have valuations of $119 billion and $150 billion and annual sales of $4.81 billion and $9.22 billion, respectively. Zscaler’s inclusion in Project Glasswing was announced weeks later and the company is smaller, with a valuation of $23 billion and sales of $2.67 billion.<\/p>\n
How Anthropic’s Opus-Only Partners Compare to Mythos Partners<\/p>\n
Those three companies are considerably better capitalized than those only invited to access Anthropic’s Opus 4.7, with SentinelOne and TrendAI parent company Trend Micro getting valuations of $5.07 billion and $4.71 billion on annual sales of $1 billion and $1.93 billion. With Opus 4.7, partners can spot issues that don’t follow predictable patterns such as logic flaws, non-linear attack paths and supply-chain risks.<\/p>\n
With Opus 4.7, Claude can analyze how code behaves across an entire system, tracing data flows and understanding interactions between components. This helps the tool surface complex vulnerabilities that may not be obvious through static pattern matching, particularly those that emerge only in specific runtime or architectural contexts. The advances are limited by not wanting to make hackers powerful.<\/p>\n
That pales in comparison to Mythos Preview, which Palo Alto used to spot complex vulnerabilities that prior-generation models missed entirely and CrowdStrike used to improve the speed and the contextual linkage of vulnerability discovery. Zscaler will integrate Claude Mythos Preview into its secure software development life cycle to uncover vulnerabilities in its software stack and Zero Trust Exchange faster.<\/p>\n
Anthropic is chasing not only product partnerships for Mythos and Opus, but also services partnerships with Accenture, BCG, Deloitte, Infosys and PwC now helping organizations deploy Claude-integrated security solutions. Major cyber vendors are also building their own AI security services squads, with Palo Alto teaming up with Accenture, Deloitte, IBM, NTT Data and PwC to propel enterprise AI resilience.<\/p>\n
CrowdStrike, meanwhile, is teaming up with two of Palo Alto’s services partners – Accenture and IBM – as well as free agents EY, Kroll and OpenAI for Project QuiltWorks, which seeks to assess, prioritize and continuously remediate the wave of vulnerabilities in production code now being discovered by frontier AI models.<\/p>\n
Anthropic has increasingly become the dominant AI lab, with its share of the enterprise LLM API market increasing from just 12% in 2023 to 32% by mid-2025, according to Menlo Ventures. This growth has come at the expense of OpenAI, which experienced a market share drop 50% to just 25% over the same period. And in the coding market, Anthropic is doubling up OpenAI, with 42% share compared to OpenAI’s 21% share.<\/p>\n
Given Anthropic’s growing importance, it isn’t surprising cybersecurity vendors big and small want to put their eggs into each product basket. But gatekeepers abound, and only the biggest and most important vendors are granted access to all the baskets.<\/p>\n
![\"Why](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/05\/anthropic-draws-line-between-who-access-opus-mythos-image_large-5-a-31588.jpg\")
<\/p>\n