Daily News: April 8, 2026

Story 1: Anthropic’s Project Glasswing — AI Finds Thousands of Critical Vulnerabilities

Anthropic has assembled a coalition of tech giants to secure the world’s most critical software using an unreleased AI model that found thousands of previously unknown vulnerabilities.

The Details:

• Apple, Google, Microsoft, Amazon, Nvidia, Cisco, JPMorgan, CrowdStrike, and the Linux Foundation have joined Project Glasswing
• The new model, Claude Mythos Preview, operates autonomously to scan code for vulnerabilities
• It found “thousands” of zero-day vulnerabilities, including a 27-year-old bug in OpenBSD and a 16-year-old flaw in widely-used video software
• One bug was in code that automated testing tools had scanned 5 million times without catching

Key Points:

• Existential threat level: Fierce competitors wouldn’t work together this way unless the alternative was mutually assured destruction of shared infrastructure
• Attack window collapse: What once took months now happens in minutes with AI, according to CrowdStrike’s CTO
• Not publicly released: Mythos won’t be made generally available — too dangerous in wrong hands
• $100M commitment: Anthropic providing $100M in Claude usage credits plus $4M in direct donations to open-source security foundations

Source: ZDNet, Anthropic

---

Story 2: Anthropic Hits $30B Revenue, Commits to 3.5GW of Google TPUs

Anthropic’s run-rate revenue has tripled from $9B (end of 2025) to $30B now, and they’ve signed a massive compute deal with Google and Broadcom.

The Details:

• 1,000 customers now spending $1M+ annually (doubled in 2 months)
• Committed to 3.5 gigawatts of Google TPU capacity starting in 2027
• Broadcom flagged Anthropic as a financial risk in regulatory filing due to the scale of commitment
• This positions Anthropic as Google’s flagship AI customer, reducing reliance on Nvidia GPUs

Key Points:

• Speed of growth: Revenue tripled in months — AI demand isn’t slowing
• Compute concentration: 3.5GW is massive infrastructure — entire data centers dedicated to one company
• TPU bet: Betting on Google’s custom chips over Nvidia GPUs for future compute

Source: The Register, SiliconANGLE

---

Story 3: OpenAI Proposes Robot Taxes and Public Wealth Funds

OpenAI released a 13-page policy document outlining how governments should handle AI-driven economic disruption, including robot taxes and a four-day workweek.

The Details:

• Document titled “Industrial Policy for the Intelligence Age” urges shifting tax burden from labor to capital
• Proposes creating a Public Wealth Fund to give Americans automatic stakes in AI companies
• Suggests subsidizing a four-day workweek with no loss in pay
• Recommends portable benefits that follow workers across jobs

Key Points:

• Tax shift: Higher taxes on corporate income, AI-driven returns, capital gains
• Robot tax concept: Robots should pay taxes equivalent to the humans they replace
• Portable benefits: Employer-subsidized healthcare and retirement that move with workers

Source: TechCrunch, The Register

---

Story 4: AI Alliance Launches Project Tapestry for Open AI Infrastructure

A coalition of 200+ organizations has launched Project Tapestry to build collaborative, sovereign AI infrastructure as an alternative to closed AI development.

The Details:

• Led by the AI Alliance, a non-profit coalition including IBM, Meta, and others
• Focus on open-source and sovereign AI capabilities
• Counter to closed AI development from OpenAI, Anthropic, Google
• Aims to ensure AI development isn’t controlled by just a few companies

Key Points:

• Sovereignty focus: Countries and organizations can build AI capabilities they control
• Open alternative: Infrastructure for those who don’t want to depend on closed AI
• 200+ members: Broad coalition from academia, industry, and government

Source: AI Alliance press release

---

Story 5: China’s Origin Wukong — Quantum AI Breakthrough

China’s 72-qubit Origin Wukong quantum computer has completed the world’s first fine-tuning of a billion-parameter AI model.

The Details:

• 76% fewer parameters needed for equivalent results
• 8.4% improvement in training outcomes
• 350,000+ quantum tasks completed since January 2024
• Users from 139 countries accessing the system

Key Points:

• World first: No other quantum computer has fine-tuned a billion-parameter model
• Efficiency gain: Using quantum computing to make AI more efficient, not just bigger
• Global access: 139 countries suggests broad access beyond China

Source: China Daily, Xinhua

---

Summary: What This Means for AI Today

April 8, 2026 was a big day for AI infrastructure. Anthropic’s Project Glasswing revealed that AI can find vulnerabilities human testing missed for decades. Meanwhile, Anthropic’s $30B revenue and massive TPU commitment show AI demand accelerating.

On the policy side, OpenAI’s economic policy document is essentially a “here’s what’s coming” warning about AI-driven disruption. And globally, we see two alternative paths: the AI Alliance’s Project Tapestry for open, sovereign AI, and China’s quantum-AI integration pushing efficiency over raw scale.

The bottom line: The AI race is no longer just about who has the best model. It’s about who secures infrastructure, who shapes policy, and who controls the compute.

Sources

• Anthropic Project Glasswing announcement
• ZDNet: “Apple, Google, and Microsoft join Anthropic’s Project Glasswing”
• The Register: “Anthropic reveals $30bn run rate”
• TechCrunch: “OpenAI’s vision for the AI economy”
• AI Alliance: “Project Tapestry launch”
• China Daily: “Origin Wukong quantum AI milestone”