Anthropic April 2026 Announcement: Everything Claude Shipped This Month

April 2026 has been the most eventful month in Anthropic's history. In seven days, the company launched a cybersecurity consortium, released a restricted frontier model that found thousands of unknown vulnerabilities, hit a $30 billion revenue run rate, closed a $380 billion valuation, shipped two major product launches, signed infrastructure deals worth billions, and started exploring custom chip design. This post covers every Anthropic April 2026 announcement in chronological order.

Key Takeaways

• Anthropic shipped seven major announcements between April 7 and April 13, 2026, spanning security, infrastructure, products, and policy
• Claude Mythos Preview found thousands of unknown zero-day vulnerabilities but Anthropic delayed public release due to offensive risk
• Revenue tripled from $9B (late 2025) to $30B, driven by 1,000+ enterprise customers each spending $1M+ annually
• New infrastructure deals with Google/Broadcom (3.5GW TPU capacity) and CoreWeave (multi-year GPU cloud) plus exploration of custom Anthropic silicon
• Claude Managed Agents and Claude Cowork reached GA, giving developers and teams production-ready agent tooling

Full Timeline: Every Anthropic April 2026 Announcement

Project Glasswing and Claude Mythos Preview (April 7)

The week opened with the biggest cybersecurity announcement in AI history. Anthropic launched Project Glasswing, a consortium of eleven organizations dedicated to using Claude Mythos for defensive security research:

1. AWS
2. Apple
3. Broadcom
4. Cisco
5. CrowdStrike
6. Google
7. JPMorganChase
8. Linux Foundation
9. Microsoft
10. NVIDIA
11. Palo Alto Networks

Claude Mythos Preview is a gated frontier model purpose-built for vulnerability discovery. During internal testing, it autonomously found thousands of previously unknown vulnerabilities across every major operating system and web browser. Anthropic restricted access to the 40+ organizations in the Glasswing consortium rather than releasing it publicly.

The decision triggered immediate debate. Bloomberg published "Mythos: Why Anthropic's New AI Has Officials Worried." Critics argued that restricting the model creates a two-tier security landscape where only large organizations benefit. Defenders countered that public release would hand offensive capabilities to attackers before patches could be developed.

By April 12, Anthropic confirmed Mythos would not receive a public release. API and web access remain restricted to Glasswing partners only.

$30 Billion Revenue and $380 Billion Valuation (April 7)

Also on April 7, Anthropic confirmed its annual revenue run rate had passed $30 billion, tripling from roughly $9 billion at the end of 2025. The company now counts over 1,000 enterprise customers each spending more than $1 million per year. Claude Code alone generates over $2.5 billion in annualized revenue.

The Series G funding round closed at a $380 billion valuation, making Anthropic one of the most valuable private technology companies in history.

Infrastructure Expansion (April 7 and April 10)

Anthropic signed two major compute deals and began exploring a third path to securing AI chip supply.

Google/Broadcom TPU Deal (April 7). Anthropic secured 3.5 gigawatts of next-generation TPU capacity fabricated by Broadcom, with delivery starting in 2027. This is one of the largest single compute commitments in AI history.

CoreWeave Multi-Year Agreement (April 10). CoreWeave announced a multi-year deal to provide NVIDIA GPU cloud infrastructure for Claude production workloads in U.S. data centers. CoreWeave stock rose 10% on the news. With this deal, nine of the ten largest AI model providers now use CoreWeave.

Custom Chip Exploration (April 10). Reuters reported that Anthropic is in the preliminary stages of evaluating in-house chip design, following the path Amazon (Trainium), Google (TPU), and Meta (MTIA) have already taken.

Claude Managed Agents and Cowork GA (April 8 to 9)

Two product launches landed back to back.

Claude Managed Agents (April 8) entered public beta. Managed Agents let developers deploy autonomous Claude agents with built-in sandboxing, state management, and the ant CLI. This removes the need to build agent infrastructure from scratch. The system supports tool use, file management, and persistent sessions out of the box.

Claude Cowork (April 9) reached general availability on macOS and Windows within Claude Desktop. Cowork lets non-technical team members work alongside Claude in a shared workspace with analytics dashboards and role-based access controls. For enterprises, it includes an analytics API for tracking agent usage and performance.

Advisor Tool (April 9) also entered public beta. The advisor tool lets a cheaper executor model (Haiku or Sonnet) consult Opus 4.6 mid-generation for difficult decisions. In benchmarks, Haiku with an Opus advisor more than doubled its SWE-bench score from 19.7% to 41.2% at 85% lower cost than running Sonnet alone.

Pentagon Dispute and Policy Developments (April 8 to 13)

Running in parallel with the product announcements, Anthropic faced ongoing tensions with the U.S. government.

On April 8, an appeals court denied Anthropic's stay request in its Pentagon contracting dispute. Anthropic remains barred from Pentagon contracts while the lawsuit over AI supply chain risk continues.

On April 11, reporting revealed that Anthropic had hosted 15 prominent Christian religious leaders at its San Francisco headquarters for two days of meetings about Claude's moral development. Topics included how Claude should respond to grief and self-harm, whether AI could be considered having moral worth, and how to embed ethical reasoning into the system.

By April 13, co-founder Jack Clark spoke at the Semafor World Economy Summit in Washington, confirming Anthropic is "talking to the administration about Mythos" and framing the Pentagon dispute as a "narrow" contracting issue separate from broader national security engagement.

What This Means for Developers

The April 2026 announcements signal several shifts for teams building on Claude:

Agent tooling is production-ready. Between Managed Agents, the ant CLI, and the Advisor tool, Anthropic now provides a complete stack for deploying autonomous agents. Teams no longer need to build sandboxing, state management, or multi-model orchestration from scratch.

Compute diversity reduces risk. With AWS Trainium, Google TPUs, CoreWeave NVIDIA GPUs, and custom chips under exploration, Claude's infrastructure is becoming more resilient. This should translate to better uptime and eventually lower prices.

Security is a first-class use case. Project Glasswing positions defensive cybersecurity as a primary Claude application. Teams building security products should monitor Glasswing for early access opportunities as the consortium expands.

Anthropic Revenue Growth: 2024 to 2026

Sources

• Reuters/CNBC: Anthropic weighs building its own AI chips
• CoreWeave: Multi-Year Agreement With Anthropic
• Bloomberg: Anthropic's New AI Model Prompts Cyber Risk Warning
• TechCrunch: Anthropic compute deal with Google and Broadcom
• TechCrunch: At HumanX, everyone was talking about Claude
• Washington Post: Anthropic consults Christian leaders on Claude's morals
• RTE: Claude Mythos public release delayed
• KFGO: Jack Clark on Trump administration and Mythos
• Fortune: Anthropic's Mythos is a wake-up call
• The Hill: Anthropic restricts release of new AI model