Tiered Access Model for AI Security Tools
Anthropic has implemented a stratified access system for its latest AI models, granting only the largest cybersecurity vendors full access to its most powerful offerings. CrowdStrike, Palo Alto Networks, and Zscaler, each valued at over $23 billion, have been designated as the exclusive launch partners for Project Glasswing, which provides access to the cutting-edge Claude Mythos Preview AI model. These three cyber titans will use Mythos to strengthen their own defenses and embed Anthropic’s most powerful generally available model, Claude Opus 4.7, into their tools.
Differences Between Opus and Mythos Access
Smaller vendors like SentinelOne and Trend Micro’s TrendAI unit have only been granted access to Claude Opus 4.7 through integrations in their tools, without access to Mythos Preview. Anthropic stated that Mythos represents an entirely new class of AI model and will not be made generally available due to the risk of adversaries using it for malicious purposes. This conservative approach contrasts with OpenAI’s more democratic model for GPT-5.4 Cyber, which uses objective criteria and identity verification to grant access to a broader range of security partners including Cloudflare and SpecterOps.
Capabilities and Market Implications
Claude Opus 4.7 can analyze how code behaves across entire systems, tracing data flows and identifying complex vulnerabilities like logic flaws, non-linear attack paths, and supply chain risks. Mythos Preview goes further, with Palo Alto using it to detect vulnerabilities that prior-generation models missed entirely. Anthropic’s enterprise LLM API market share has grown from 12% in 2023 to 32% by mid-2025, while OpenAI’s share dropped from 50% to 25%. No specific CVEs were mentioned in this article.
Source: Healthcareinfosecurity
