AI Furnace Newsroom: ​​Anthropic’s Mythos Backlash, U.S. AI Labs Team Up to Halt the Challenge from China, Claude Cowork Goes From Experiment To Rollout Product

Welcome to this week’s AI Furnace Newsroom

In today’s insights we cover:

1. Anthropic’s Mythos Backlash: Project Glasswing Is Anthropic’s Attempt To Get Ahead Of A Cybersecurity Shock

2. U.S. AI Labs Team Up to Halt the Challenge from China

3. Claude Cowork Goes From Experiment To Rollout Product

Read time: 5 mins

💡 Furnace Insights

Anthropic

Anthropic’s Mythos Backlash: Project Glasswing Is Anthropic’s Attempt To Get Ahead Of A Cybersecurity Shock

Anthropic’s new Project Glasswing is built around Claude Mythos, a model the company says is powerful enough to find and exploit serious software vulnerabilities across major operating systems and browsers. Anthropic is not releasing the model publicly. Instead, it is giving selected partners early access for defensive use, backing the effort with up to $100 million in usage credits and $4 million in donations to open-source security groups. Launch partners include AWS, Apple, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, Nvidia, Palo Alto Networks, and the Linux Foundation.  

This is one of the clearest signs yet that frontier AI labs think cyber capability is moving into a different league. Anthropic is basically saying: this model is useful enough to defend with, but risky enough that releasing it broadly would be irresponsible. That is a striking position. It also comes with a clock attached, because Anthropic says comparable models from others may arrive within six to eighteen months. In other words, Glasswing is less a victory lap than a head start.

AI Market

U.S. AI Labs Team Up to Halt the Challenge from China

OpenAI, Anthropic, and Google are reportedly sharing intelligence through the Frontier Model Forum to detect and block adversarial distillation, where outside actors hammer frontier APIs with automated queries to clone model behavior without paying the original training cost. The effort reportedly follows a series of accusations from Anthropic that DeepSeek, Moonshot, and MiniMax extracted Claude capabilities at scale, and it reflects growing concern in Washington as well as Silicon Valley.  

The interesting part is not just the anti-China angle. Commercial rivals are now treating model outputs as strategic infrastructure worth defending together. That tells you distillation is no longer seen as ordinary competition or a nuisance abuse problem. It is becoming a direct challenge to the economics of frontier AI. If top labs think copied capability can erase billions in R&D advantage, expect much tighter account verification, more aggressive monitoring, and less openness around high-end access. 

Anthropic

Claude Cowork Goes From Experiment To Rollout Product

Anthropic is pushing Claude Cowork into company-wide deployment mode. It is now available on paid plans across macOS and Windows, and the new admin layer adds the kind of controls enterprises usually wait for before saying yes: role-based access, team budgets, usage analytics, OpenTelemetry support, and connector-level permissions. Anthropic is also adding a Zoom connector so meeting summaries and action items can flow straight into Cowork workflows. 

Anthropic says a lot of Cowork usage is coming from outside engineering, which is a quiet but important shift. AI agents are no longer just a developer tool story. They are becoming a finance, legal, ops, and marketing story too. The real unlock is not “better AI,” it is governance that makes broad adoption tolerable for IT, security, and budget owners.

📈 AI Venture Deals of the Week

• Aria Networks, an AI-native data center networking infrastructure startup, raised its $125M Series A. 

• PeakMetrics, an AI-powered narrative intelligence and media monitoring platform, raised its $6M Series A. 

• Juno, an AI tax preparation automation platform for accounting firms, raised its $12M Seed. 

• Nas, an AI platform for solopreneurs and one-person businesses, raised its $27M Series A.

• Narwhal Labs, an autonomous AI communications platform for regulated industries, raised its £20M funding round.

⚒️ New AI Product Launches You Don’t Want to Miss 

• Meta released Muse Spark, a flagship model powering Meta AI across its apps. 

• Resend, a developer communications infrastructure platform, released its official MCP Server for AI agents. 

• Knot, a transaction and subscription data infrastructure platform, released an AI-powered exploration agent inside its dashboard. 

• Arcee, a open model infrastructure company, released Trinity-Large-Thinking, an open reasoning model for long-horizon AI agents. 

• Z.ai, a multimodal foundation model platform, released GLM-5V-Turbo, a vision-language coding and agent execution model.

Upcoming Events 📅

• AIF Founder Supper Club @ Private Room - Tuesday, April 14

Interested in meeting the Who’s Who of AI? Join the AI Hot 100 Summit in NYC

The AI Hot 100 Summit is back for its third edition on May 6-7th, 2026 in New York City. The only AI Summit built for real connections where AI Visionaries meet Industry Leaders. Join over 500 AI Executive Leaders, Founders and Investors for a jam packed 2 days of enterprise AI lightning talks, innovation showcases, and curated networking.

Get ready to meet and greet speakers and attendees from OpenAI, NVIDIA, Haleon plc, Legora, Anthropic, Cursor, Eleven Labs, Adobe, Walmart, Insight Partners & more.

🗓️ May 6-7, 2026

📍New York City