Hello Builders, issue #13
News from the trenches of AI in the week Feb 17-23, 2026
Hello Builders,
This was the week the AI cold war went hot. Anthropic publicly accused DeepSeek, Moonshot, and MiniMax of industrial-scale distillation — 24,000 fake accounts, 16 million exchanges, proxy networks to evade detection. Not a rumor, not a leak: a formal accusation framed as national security. OpenAI followed with its own allegations to Congress. Meanwhile, Google did something that may matter more than any model release this year: Chrome 146 shipped with native WebMCP support, turning every website into a potential AI agent endpoint. OpenAI launched Frontier, an enterprise platform for “AI coworkers,” and signed alliances with McKinsey, BCG, Accenture, and Capgemini to deploy it. Meta signed a multi-billion-dollar GPU deal with Nvidia covering millions of chips. And every major AI CEO flew to Delhi for India’s AI Summit, where billions in deals were announced in four days. The pattern is unmistakable: AI is no longer a technology race. It’s an infrastructure war, a trade war, and a regulatory war — all at once.
This week’s signal in the noise
• Anthropic accuses DeepSeek, Moonshot, and MiniMax of running 24,000 fake accounts to distill Claude’s capabilities at an industrial scale
• Google ships WebMCP in Chrome 146, creating a native protocol for AI agents to interact with any website
• OpenAI launches Frontier, an enterprise platform for “AI coworkers,” and signs multi-year alliances with McKinsey, BCG, Accenture, and Capgemini
• Meta expands Nvidia partnership to deploy millions of GPUs and next-gen Vera Rubin systems in a deal worth tens of billions
• India AI Summit draws Altman, Amodei, Pichai, and Huang, with billions in infrastructure commitments announced
1. Anthropic vs. China: 24,000 Fake Accounts and a National Security Accusation
Anthropic dropped a bombshell on Sunday: DeepSeek, Moonshot AI, and MiniMax ran coordinated campaigns using 24,000 fraudulent accounts and proxy networks to extract Claude’s outputs and train their own models. Anthropic generated 16 million exchanges across these campaigns and called it “industrial-scale.” DeepSeek specifically sought to create “censorship-safe alternatives to policy-sensitive queries.” Anthropic framed the disclosure as a national security issue, timed to coincide with Washington's debate over AI chip export controls. OpenAI had made similar allegations about ChatGPT to the House Select Committee earlier this month. When the two fiercest competitors in AI agree on something, the signal is deafening. For builders: distillation defense is now a first-class engineering concern. If you’re serving frontier model outputs via API, your responses are a target.
2. Google Ships WebMCP: The Most Important Release No One Is Talking About
Chrome 146 shipped with native WebMCP (Web Model Context Protocol) support — a standard that lets AI agents interact with websites via structured “Tool Contracts” rather than fragile screen scraping. Websites can now declare what actions agents are allowed to take: book a flight, add to cart, submit a form. The browser becomes the universal agent runtime. If WebMCP gains adoption, every website effectively becomes an MCP server. This is what the “agentic web” actually looks like — not chatbots slapped onto homepages, but a fundamentally new interaction layer where AI negotiates with websites on your behalf. For builders: start implementing WebMCP endpoints now. The companies that expose structured agent interfaces first will capture the agentic traffic wave that’s building.
3. OpenAI Launches Frontier: AI Coworkers Get Onboarded Like Employees
OpenAI launched Frontier, an enterprise platform for building, deploying, and managing “AI coworkers” — complete with onboarding, permissions, identity governance, and audit trails. Think of it as HR for agents. Then, on Sunday, OpenAI signed multi-year alliances with McKinsey, BCG, Accenture, and Capgemini to deploy Frontier across their enterprise clients. Brad Lightcap said the quiet part out loud: “The limiting factor for seeing value from AI in enterprises isn’t model intelligence — it’s how agents are built and run in their organizations.” This is OpenAI’s boldest enterprise move yet. They’re not selling APIs anymore; they’re selling a managed agent workforce. The consulting partnerships are the delivery mechanism—the Big Four serve as OpenAI’s implementation layer. For builders: if McKinsey and Accenture are deploying OpenAI’s agents, they’re coming for enterprise workflows at scale. The SaaS disruption everyone predicted is now being actively orchestrated.
Link: https://www.cnbc.com/2026/02/23/open-ai-consulting-accenture-boston-capgemini-mckinsey-frontier.html
4. Meta Goes All-In: Millions of Nvidia GPUs, $130B AI Budget
Meta expanded its Nvidia partnership in a deal worth tens of billions, deploying millions of GPUs — including Nvidia’s new standalone CPUs and next-generation Vera Rubin systems — across its data center fleet. Jensen Huang called it “bringing the full NVIDIA platform to Meta.” The company’s total AI spend for 2026: $115-135 billion across 30 new data centers. Meta AI already has 1 billion users. Zuckerberg is betting that whoever controls the most compute wins the consumer AI race. Combined with Meta’s open-source strategy through Llama, this creates a unique position: the largest compute footprint in AI, powering the largest open-weight model ecosystem. For builders: the open-weight ecosystem just got a massive infrastructure boost. Plan for a world where the best open models are genuinely competitive with proprietary ones.
Link: https://www.cnbc.com/2026/02/17/meta-nvidia-deal-ai-data-center-chips.html
5. India AI Summit: Every CEO at Modi’s Court
Every major AI leader flew to New Delhi this week. Sam Altman, Dario Amodei, Sundar Pichai, and Jensen Huang all attended the four-day AI Impact Summit and met directly with Prime Minister Modi. The commitments poured in: L&T partnered with Nvidia for AI data center infrastructure, India’s payments network adopted Nvidia’s Nemotron for sovereign AI, OpenAI struck a platform deal with Zomato, and Google DeepMind announced a partnership with the Indian government for science and education. Pichai publicly defended AI spending against bubble fears. The message was unanimous: India isn’t just a market — it’s the next strategic battleground, sitting between the US and China with 1.4 billion people and accelerating digital infrastructure. For builders: the global south is where the next billion AI users come from. Build for multilingual, mobile-first, cost-sensitive deployment.
Link: https://techcrunch.com/2026/02/22/all-the-important-news-from-the-ongoing-india-ai-summit/
6. The Year of the AI Bill: US States Launch a Regulatory Flood
While the industry raised money and fought geopolitical battles, US states launched an unprecedented wave of AI legislation. Alabama signed the App Store Accountability Act, targeting digital child safety. Idaho’s Conversational AI Safety Act advanced through committee with a do-pass recommendation. Colorado’s comprehensive AI Act takes effect June 30, requiring “reasonable care” to prevent algorithmic discrimination. At the federal level, Anthropic and OpenAI’s distillation accusations are feeding the push for stricter export controls. The EU AI Act’s prohibited practices provisions are already in force. The regulatory landscape is no longer hypothetical — it’s operational. For builders: if you’re deploying AI in production, you need a compliance strategy now, especially for high-risk use cases in hiring, lending, healthcare, and any consumer-facing application.