Project Glasswing: 12 Tech Giants Unite on AI Security

Twelve of the world's most powerful tech companies — including direct competitors Apple, Google, and Microsoft — just signed on to share a security table with Anthropic. The initiative is called Project Glasswing, and its mission is to protect the software that critical infrastructure (power grids, banking systems, hospital networks) depends on. Simultaneously, Anthropic released findings from an 81,000-person user study that has already produced one permanent policy: Claude will never carry ads.

Project Glasswing: How 12 Tech Rivals Are Cooperating on AI Security

Project Glasswing's founding coalition reads like a who's who of companies that normally compete directly:

• AWS and Google — the two largest cloud computing providers, fighting for the same enterprise contracts
• Microsoft — which operates its own AI security products and Copilot integrations
• Apple — which almost never participates in cross-industry security initiatives of this kind
• NVIDIA — the dominant AI chip manufacturer
• CrowdStrike, Cisco, Palo Alto Networks — three competing cybersecurity firms
• JPMorganChase — representing the financial services sector, one of the highest-value cyberattack targets globally
• Linux Foundation — the neutral steward of much of the world's open-source software (code anyone can use, modify, and share for free)
• Broadcom — semiconductor and infrastructure software leader

The shared problem is real: critical infrastructure software (the code running power grids, water systems, hospital records, and financial networks) is an increasing target for nation-state actors and sophisticated criminal groups. No single company — not even Microsoft or Google at their scale — has full visibility into vulnerabilities across the entire software supply chain (the interconnected web of open-source libraries, commercial frameworks, and cloud services that modern software depends on).

Detailed implementation timelines and governance structures for Project Glasswing haven't been disclosed yet. The Linux Foundation's role as a neutral venue is significant: it means no single company controls the standards being set, and the open-source ecosystem — which underpins the software security of virtually every major organization — gets a seat at the table alongside enterprise players.

Claude AI Goes Ad-Free: What 81,000 Users Told Anthropic

Anthropic has been running what it describes as the largest and most multilingual qualitative study (research collecting open-ended answers rather than multiple-choice responses) ever conducted about AI usage. The sample: 81,000 Claude.ai users across languages and regions — roughly 8 to 16 times larger than the typical consumer tech survey of 5,000–10,000 participants.

Three questions drove the research:

1. How do you use AI in your daily life?
2. What do you aspire for AI to eventually help you with?
3. What concerns do you have about AI?

Full findings haven't been released. But one concrete policy decision is directly tied to this research: Claude will remain permanently ad-free.

Anthropic's stated reasoning is direct: advertising incentives are "incompatible with a genuinely helpful AI assistant." The logic cuts to the business model level (how a company actually makes its money). An ad-supported AI is structurally (by design of its revenue model) incentivized to keep users engaged and browsing, not to solve their problem in 30 seconds and let them go. An AI that prolongs conversations earns more ad revenue. An AI that solves your problem immediately does not.

Several competitors are already exploring personalized ad targeting using conversation data. If you use Claude for work, this distinction matters in practice: your conversation data is not being used to build an advertising profile — a meaningful difference if you discuss client work, internal strategy, or personal finances with the tool.

Anthropic's $100M AI Partner Network and Global Expansion

The two headline announcements sit inside a broader expansion that's unusual in its speed and scope. In the same period, Anthropic has:

• Invested $100 million into the Claude Partner Network — an ecosystem fund for companies building products on top of Claude
• Secured a compute (raw AI processing power) expansion deal with Google and Broadcom covering multiple gigawatts of next-generation infrastructure
• Established Sydney as its fourth Asia-Pacific office
• Signed a formal Memorandum of Understanding (an official collaboration agreement) with the Australian government on AI safety and research
• Launched a partnership with Mozilla to strengthen Firefox browser security
• Founded the Anthropic Institute as a dedicated AI research entity
• Appointed Vas Narasimhan — CEO of pharmaceutical giant Novartis — to its Long-Term Benefit Trust Board of Directors

The $100 million partner fund is aimed directly at developers. If you're currently building tools on Claude, Anthropic's partner ecosystem is actively expanding with funding attached. The practical effect: more third-party products built on Claude creates more enterprise lock-in — a competitive moat (barrier to switching) measured in developer ecosystem size, not just benchmark scores.

Claude Sonnet 4.6 and Opus 4.7: New AI Automation Models for Enterprise

The week's announcements coincide with two model updates:

• Claude Sonnet 4.6 — positioned for "frontier performance" (state-of-the-art results) across coding tasks, AI agents (automated systems that complete multi-step workflows), and professional use at organizational scale
• Claude Opus 4.7 — Anthropic's highest-capability model tier, updated for this product cycle

For coding-heavy work or setting up AI automation workflows, Sonnet 4.6 is available now at Claude.ai. For organizations evaluating AI vendor risk: Project Glasswing's formation signals that infrastructure-level software security is becoming a cross-industry standard — not a single-vendor proposition. The companies that control critical software pipelines (deployment systems, cloud infrastructure, open-source dependencies) are now aligning on shared standards, and AI tools embedded in those pipelines will eventually be expected to meet them. Watch Glasswing's governance structure once it's published — it will define those expectations concretely.

Related Content — Get Started | Guides | More News