Anthropic @ $30B ARR, Project GlassWing and Claude Mythos Preview — first model too dangerous to release since GPT-2

2026-04-07 · Source: AINews · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, Robotics & Autonomous Systems · Depth: Advanced, extended

Summary

Anthropic has significantly escalated its competitive stance against OpenAI, reporting a jump from $19B ARR in March to $30B ARR in April 2026, amidst OpenAI's $24B ARR announcement and stalled ChatGPT growth. Concurrently, Anthropic formally confirmed "Claude Mythos," a large language model rumored to be the largest successful training run, which they deem too dangerous for general release. Instead, Mythos is restricted to 40 partners under "Project Glasswing," an urgent cyberdefense initiative. This model demonstrates shocking capabilities, including finding thousands of high-severity vulnerabilities in major operating systems and web browsers, with benchmark scores like 93.9% on SWE-Bench Verified and 100% on Cybench CTF. The move signals a strategic shift towards withholding frontier models for controlled, high-stakes applications, supported by Anthropic's rapidly compounding revenue.

Key takeaway

For CTOs and VPs of Engineering evaluating AI adoption strategies, Anthropic's decision to restrict Claude Mythos to Project Glasswing partners signals a new era where the most powerful AI capabilities may not be publicly accessible. You should assess your organization's cybersecurity posture and consider engaging with private frontier AI initiatives, as these models offer unparalleled vulnerability detection but also introduce novel risks requiring specialized governance and deployment strategies.

Key insights

Frontier AI models are increasingly being strategically withheld for high-stakes applications, driven by advanced capabilities and robust revenue.

Principles

• High-margin enterprise workloads can sustain frontier AI development without broad public access.
• Advanced AI capabilities necessitate restricted release and dual-use governance frameworks.

Method

Anthropic's Project Glasswing involves deploying Claude Mythos, a highly capable model, to a coalition of 40 partners for cyberdefense, supported by extensive safety documentation and controlled access.

In practice

• Utilize 2-bit quantization for large models like GLM-5.1 to enable local runs on 256GB RAM machines.
• Prioritize context management and agentic harnesses for practical agent improvement over raw model swaps.

Topics

• Claude Mythos
• Project Glasswing
• AI Cybersecurity
• Anthropic Revenue Growth
• LLM Benchmarks

Code references

• sandroandric/AgentHandover
• agents-io/PokeClaw
• GitFrog1111/badclaude

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Tech Journalist, Investor, AI Scientist

Related on AIssential

• Anthropic Teams Up With Its Rivals to Keep AI From Hacking Everything — Anthropic launched Mythos Preview and Project Glasswing to address advanced AI capabilities and their cybersecurity implications.
• [AINews] Anthropic @ $30B ARR, Project GlassWing and Claude Mythos Preview — first model too dangerous to release since GPT-2 — Anthropic's Claude Mythos model demonstrates unprecedented cybersecurity capabilities, leading to a restricted release for critical infrastructure defense.
• Anthropic's new AI is too powerful for the world — Anthropic's unreleased Mythos AI, deployed via Project Glasswing, demonstrates extreme power in cybersecurity, prompting restricted access.
• Anthropic's Project Glasswing - restricting Claude Mythos to security researchers - sounds necessary to me — Advanced LLMs are now autonomously finding and exploiting critical software vulnerabilities at an unprecedented scale.
• Anthropic invites 150 more organizations into Project Glasswing — Claude Mythos is an AI model capable of advanced software vulnerability detection and exploitation, raising significant security concerns and regulatory attention.
• Claude Mythos #2: Cybersecurity and Project Glasswing — Anthropic's Claude Mythos demonstrates unprecedented autonomous cyber exploitation capabilities, prompting a defensive, limited release strategy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AINews.