The New AI Access Hierarchy
Anthropic is implementing a tiered access model for its most advanced artificial intelligence models, creating a clear divide among cybersecurity vendors. At the top of the pyramid, only three companies CrowdStrike, Palo Alto Networks, and Zscaler have been granted access to the cutting-edge Claude Mythos Preview AI model through Project Glasswing. This exclusive group plans to use Mythos Preview to strengthen their own defensive capabilities and embed the model into their security tools.
One tier below, vendors such as SentinelOne and TrendAI (Trend Micro’s enterprise unit) can integrate the less powerful but still advanced Claude Opus 4.7 model into their products. However, they do not have access to Mythos Preview. Anthropic has stated it does not plan to make Mythos generally available due to the risk that adversaries could misuse it for harmful purposes.
Impact on Security Tool Capabilities
With access to Opus 4.7, partners can identify vulnerabilities that do not follow predictable patterns, including logic flaws, non-linear attack paths, and supply-chain risks. Claude with Opus 4.7 can analyze how code behaves across an entire system, tracing data flows and understanding interactions between components. This helps surface complex vulnerabilities that may not be obvious through static pattern matching, particularly those that emerge only in specific runtime or architectural contexts.
Mythos Preview offers even more advanced capabilities. Palo Alto Networks used it to spot complex vulnerabilities that prior-generation models missed entirely. CrowdStrike leveraged it to improve the speed and contextual linkage of vulnerability discovery. Zscaler plans to integrate Mythos Preview into its secure software development lifecycle to uncover vulnerabilities in its software stack and Zero Trust Exchange faster.
Market Dynamics and Partner Ecosystem
Anthropic’s tiered approach contrasts with OpenAI’s more open strategy for GPT-5.4-Cyber, which grants access to a broader group including Cloudflare, iVerify, SpecterOps, and others based on clear criteria and identity verification. Unlike Mythos, which Anthropic describes as an entirely new class of AI model, GPT-5.4-Cyber is a version of OpenAI’s publicly available GPT-5.4 with modified refusal boundaries for cybersecurity work.
Anthropic’s enterprise LLM API market share has grown from 12% in 2023 to 32% by mid-2025, according to Menlo Ventures, at OpenAI’s expense. In the coding market, Anthropic now holds 42% share compared to OpenAI’s 21%. The company is also building services partnerships with Accenture, BCG, Deloitte, Infosys, and PwC to help organizations deploy Claude-integrated security solutions. Major cyber vendors like CrowdStrike and Palo Alto Networks are forming their own AI security service teams, with CrowdStrike launching Project QuiltWorks alongside Accenture, IBM, EY, Kroll, and OpenAI to address vulnerabilities discovered by frontier AI models.
Source: Healthcareinfosecurity
