🐺 The Pentagon wants to blacklist Anthropic

Hey there,

This week in AI: the Pentagon threatened to blacklist the only AI company pushing back on unrestricted military use, OpenAI is closing a $100 billion funding round, and ChatGPT officially has ads now.

Also, the guy who built AlphaGo just left Google DeepMind and is raising $1 billion to build AI that doesn't use large language models at all. If that doesn't make you rethink the current playbook, nothing will.

Let's get into it.

📰 WHAT'S TRENDING

The Pentagon Is Ready to Blacklist Anthropic Over Safety Guardrails Defense Secretary Pete Hegseth is reportedly "close" to cutting business ties with Anthropic and designating it a "supply chain risk" — a label typically reserved for foreign adversaries. The issue: Anthropic wants guardrails preventing Claude from being used for mass surveillance of Americans or autonomous weapons. The DOD wants unrestricted access for "all lawful use cases." OpenAI, Google, and xAI have all agreed to lift their restrictions. Anthropic is the only holdout — and it's the only AI company currently deployed on classified military networks. 🔗 Read more

OpenAI Finalizing $100B Round at $850B+ Valuation Amazon may invest up to $50B, SoftBank around $30B, and Nvidia roughly $20B. Microsoft is also in. The pre-money valuation sits at $730B, with post-money clearing $850B. First close is expected before the end of February. OpenAI is also reportedly considering going public as early as late 2026. For context: this single funding round is larger than the GDP of most countries. 🔗 Read more

ChatGPT Now Has Ads Ads started rolling out February 9 to Free and Go tier users in the U.S. Brands testing include Target, Adobe, Ford, HelloFresh, and Audemars Piguet. They appear as labeled "sponsored" boxes matched to conversation topics. Plus, Pro, Business, and Enterprise tiers remain ad-free. With 200M+ weekly users, ChatGPT just became a major advertising platform. Sam Altman, who once called combining ads with AI "uniquely unsettling," apparently got over it. 🔗 Read more

Alibaba Drops Qwen 3.5 — Claims It Beats GPT-5.2 and Claude Opus 4.5 397 billion parameters. Native multimodal (text, images, video). Open-weight. Supports 201 languages. Alibaba says it outperforms GPT-5.2, Claude Opus 4.5, and Gemini 3 — though independent benchmarks haven't confirmed this yet. It's 60% cheaper to run and has 8x more throughput than the previous version. If the claims hold, a free model just matched the best closed-source competitors. 🔗 Read more

The AlphaGo Creator Left Google DeepMind to Bet $1B on Reinforcement Learning David Silver — the researcher behind AlphaGo — founded Ineffable Intelligence in London and is raising $1 billion at a $4B pre-money valuation. Sequoia is leading. Nvidia, Google, and Microsoft are reportedly in talks to participate. The bet: reinforcement learning, not LLMs, is the path to superintelligence. While everyone else scales transformers, Silver is building something fundamentally different. 🔗 Read more

Anthropic Ships Claude Sonnet 4.6 with 5x Better Computer Use Sonnet 4.6 scores 72.5% on the OSWorld computer use benchmark — nearly a 5x improvement in 16 months. It handles multi-step web forms, complex spreadsheets, and office tasks that previously required the Opus model. 1M token context window. Priced at $3/$15 per million tokens. Now the default model for Free and Pro users. 🔗 Read more

Meta and Nvidia Lock in a Multi-Year AI Infrastructure Deal Meta is the first company to deploy Nvidia's Grace CPUs as standalone chips in data centers. This comes on top of Meta's January announcement: $135 billion in AI spending for 2026. One company. One year. $135 billion. The AI infrastructure arms race is not slowing down. 🔗 Read more

🧠 HERE'S THE THING

The Pentagon is threatening to blacklist Anthropic — the same company that holds a $200M DOD contract and runs on classified military networks because it won't remove safety guardrails.

Meanwhile, OpenAI, Google, and xAI all agreed to the DOD's terms with no restrictions and full access.

Think about what that means. Three of the four major AI companies said "yes sir" to unrestricted military use. The one company that pushed back is being treated like a hostile foreign supplier.

And on the same week, Anthropic-backed PAC spent $450K supporting a congressional candidate who favors AI safety regulation.

This is the first real test of whether the AI industry can hold any ethical lines when government contracts are on the table.

The companies that built these models are deciding right now who gets to use them and how. Those decisions will matter long after the funding rounds are forgotten.

💡 Smart Moves

Cold Truth: B2B Lead Gen That Works

If you run or work at a B2B company and you need more pipeline, I write a second newsletter called JX Creative.

Twice a week. Tactical. Free thoughts. Cold email infrastructure, copywriting best practices, and how to build lists for your target audience.

Everything is based on real campaigns we've run. Tap in if you’re running or working at a B2B business.

⏳ CONVERSION CORNER

David Silver built AlphaGo. Spent 12 years at Google DeepMind. Then he quit. His new company, Ineffable Intelligence, is raising $1 billion. The thesis: LLMs are a dead end. Reinforcement learning is the real path to superintelligence.

Everyone else is spending hundreds of billions to make language models bigger. Silver thinks they're all wrong. He wants to build AI that learns from experience the way AlphaGo learned Go, not by memorizing the internet.

Maybe he's right. Maybe he's not. But the guy who solved Go isn't usually the guy you bet against.

Now it's your move: What's the "obvious" strategy in your industry that everyone's copying? Find the person with a decade of experience who walked away from it. Ask them why. That's where the next real opportunity usually lives.

💎 DATA GEM

66% of people on Earth have used an AI tool in the past year. In 2023, it was 38%. In 2024, 48%. Now two out of three humans have tried it.

Smartphones took a decade to hit those numbers. AI did it in three years.

That's it for this week. Stay sharp out there.

Jake

P.S. If this was useful, forward it to someone who works in defense tech or government AI procurement. They're going to want to read the Anthropic story before their next meeting.