AI News Summary — Week 15.2, 2026

Happy Friday! While earlier this week covered Anthropic's explosive $30 billion revenue run rate, this edition focuses on their latest announcement: Mythos, a powerful new frontier model purpose-built for cybersecurity.

Anthropic Unveils Mythos: Frontier Model for Defensive Security

Anthropic has released a preview of Mythos, a new frontier model that the company describes as among its most powerful yet. However, this isn't a typical model release — Mythos won't be available to the general public.

Instead, the model is being deployed exclusively through Project Glasswing, a new security initiative where 12 partner organizations will use Mythos specifically for "defensive security work" to secure critical software systems.

Partner Organizations

The partners include major tech and security companies:

• Amazon, Apple, Broadcom, Cisco, CrowdStrike
• Linux Foundation, Microsoft, Palo Alto Networks

In total, 40 organizations will gain access to the Mythos preview, though the general release is not planned.

Early Results: Thousands of Zero-Day Vulnerabilities Found

The results have been impressive. Anthropic reports that over the past few weeks, Mythos has already identified thousands of zero-day vulnerabilities, including many critical ones — some dating back one to two decades.

While Mythos wasn't specifically trained for cybersecurity, the model has strong agentic coding and reasoning skills that make it effective at scanning both first-party and open-source software systems for code vulnerabilities.

From Leak to Launch

The existence of Mythos was originally revealed through a data security incident last month, when a draft blog post about the model (then codenamed "Capybara") was left in an unsecured cache. The leaked document described the model as "by far the most powerful AI model we've ever developed."

This wasn't the only security mishap for Anthropic recently. The company also accidentally exposed nearly 2,000 source code files and over half a million lines of code via a mistake in the launch of Claude Code 2.1.88 — and then accidentally caused thousands of GitHub repositories to be taken down while trying to clean up the mess.

Ongoing Federal Discussions

Anthropic says it has engaged in "ongoing discussions" with federal officials about Mythos — though these conversations are complicated by the ongoing legal battle between Anthropic and the Trump administration. The Pentagon previously labeled Anthropic a "supply-chain risk" after the company refused to allow autonomous targeting or surveillance of U.S. citizens.

What's Hot This Week

• Frontier models go defensive — Mythos represents a new paradigm: powerful AI released selectively for security purposes
• AI as a security tool — finding thousands of real vulnerabilities shows AI's potential in defensive security
• Responsible AI release — limiting access to vetted partners is a contrast to the typical "release everywhere" approach
• Anthropic's month — between the data leaks and the Pentagon battle, it's been a turbulent April

That's a wrap for this week's AI news. Have a great weekend!

Sources: TechCrunch, Anthropic Blog, Axio, Fortune