Elon Musk Is Suing OpenAI for $130B

Welcome back apprentices! 👋

Hey {{first_name|friend}},

You know that feeling when you step away from your phone for a weekend and come back to 47 unread messages? 

That's basically what happened in AI this week — except instead of group chat drama, it was billion-dollar lawsuits, secret government contracts, and a Chinese AI model that nobody saw coming. 

The kind of week where every story sounds made up until you realize it isn't. 

We sorted through all of it, so here's everything you actually need to know.

_{Read Time: 5 minutes}

🕵️ Copy-Paste or Genius? The White House just dropped a memo accusing Chinese AI firms — including DeepSeek, Moonshot, and MiniMax — of running "industrial-scale" operations to steal U.S. frontier AI knowledge, using thousands of fake accounts and jailbreaks to train their own models on American AI outputs. Think of it as downloading someone else's brain instead of growing your own. The timing is spicy: this lands just weeks before Trump sits down with Xi Jinping in Beijing, and a new House bill could slap distillation offenders onto the U.S. export blacklist.

🔬 AI Sees Cancer 3 Years Early. Mayo Clinic's AI model REDMOD can now detect pancreatic cancer up to three years before a human doctor would catch it — by reading invisible texture and tissue patterns in routine CT scans that patients are already getting anyway. In a landmark study of nearly 2,000 scans that specialists had originally cleared as normal, REDMOD flagged 73% of early cases, and at the two-year mark outperformed experienced radiologists by roughly 3 to 1. No new tests, no extra steps — just a smarter second set of eyes on existing data.

🧬 AI That Teaches Itself. David Silver — the DeepMind legend behind AlphaGo — just launched Ineffable Intelligence with a jaw-dropping $1.1B seed round (Europe's largest ever) to build AI that learns purely from experience, no human data required. While most AI labs race to hoover up more training data, Silver's calling that approach "fossil fuel" and betting his "superlearner" can teach itself indefinitely through simulation alone — with the stated goal of making "first contact with superintelligence."

Manual marketing had a good run.

But the teams winning right now aren't briefing, approving, and repeating. They're directing AI agents that execute the whole strategy for them.

The Agentic Marketing Summit (May 4–8) is a free, five-day event that shows you exactly how it works in practice. Not theory. Not a PDF checklist. Step-by-step insight to help you become an expert in AI marketing agents.

Hosted by 3x Inc 5000 founder Manick Bhan alongside the sharpest minds in the marketing world today.

The era of doing it yourself is over!

Join Free Today!

If this past week felt like a normal week in AI, you clearly weren't paying attention. 

The industry didn't just move — it lurched, stumbled, sued, signed deals in secret, and raised money like it was going out of style. 

Courts, boardrooms, Pentagon offices, and Beijing regulators all had a front-row seat in the same five-day window, and the decisions made in each of those rooms are now on a collision course with each other.

This was exactly the kind of week that makes the next twelve months harder to predict — and a lot more interesting to watch.

It's been a big week for Sam Altman — though probably not a relaxing one. 

On the product side, OpenAI dropped GPT-5.5, internally codenamed "Spud" (yes, really), and it's impressive. The model tops public benchmarks in reasoning, agentic tasks, computer use, and coding — with scores nudging up against Claude Mythos territory — while running at GPT-5.4's speed and coming in at $5/$30 per million tokens. 

OpenAI is calling it "half the cost of competitive frontier coding models." After months of Anthropic riding high and OpenAI fielding criticism about quality degradation and rate limits, Spud is a statement.

Simultaneously, OpenAI rewrote the terms of its foundational partnership with Microsoft. 

The new deal kills Microsoft's exclusivity over OAI's IP, removes the vague and contentious "AGI clause" (which could have triggered partnership terms based on whenever someone decided superintelligence had arrived), and frees OpenAI to deploy on any cloud — including Amazon Bedrock. Microsoft keeps Azure-first launch access through 2032 and a revenue share through 2030, while stopping its own payments to OpenAI. 

Translation: OpenAI graduates from Microsoft's walled garden while Microsoft turns the relationship into a clean, time-bound business contract. Andy Jassy at Amazon called the news "very interesting." We'd bet he had a bigger smile than he let on.

And then there's the courtroom. Elon Musk took the stand in federal court as his $130B lawsuit against OpenAI kicked off in earnest. Musk is seeking damages, the removal of Altman and Greg Brockman from the board, and a forced reversal of OpenAI's for-profit conversion — arguing that Altman "stole a charity." OpenAI's legal team fired back, calling it "sour grapes" from someone who didn't like losing, and Microsoft's lawyers noted that Musk only started objecting to OpenAI's structure once xAI became a direct competitor. 

Day 1 set the tone: personal, dramatic, and loaded with private messages about to become public. The next four weeks are going to be required reading.

Google's week was a study in corporate contradiction. 

The company quietly signed a classified AI deal with the Pentagon — reportedly opening its models to "any lawful government purpose" with no right to veto usage — the very same week that more than 600 employees published an open letter to CEO Sundar Pichai demanding he refuse classified military workloads. 

For context: Google originally adopted a no-weapons AI policy in 2018 after staff pressure, then scrubbed it from its AI principles entirely in 2025. 

OpenAI and xAI both signed Pentagon deals last month; Anthropic is currently in court after being blacklisted for refusing to drop its guardrails. The U.S. government wants frontier AI in its toolkit, and one by one, the major labs are signing on. 

Whether the employee backlash grows into something more disruptive — like the 2018 Project Maven protests did — remains the story to watch.

DeepSeek quietly dropped preview versions of V4, and it's a competitive shot across the bow. 

V4 Pro features a 1M-token context window, sits near the top of open-model benchmarks (with DeepSeek's own evals placing it near GPT-5.5 and Gemini 3.1-Pro on reasoning), and costs just $1.74/$3.48 per million input/output tokens — a fraction of GPT-5.5's $5/$30. 

More geopolitically significant: the model runs on Huawei's Ascend chips, offering the most concrete proof yet that China can build capable AI infrastructure entirely outside Nvidia's stack. Export restrictions were supposed to slow this down. DeepSeek's pricing makes the AI race as much about cost efficiency as raw capability.

Meanwhile, China made its most aggressive move yet in the AI talent war by vetoing Meta's $2B acquisition of Manus, the Singapore-based AI startup with Chinese roots. China's NDRC stepped in, citing foreign investment and export-control violations, and directed both companies to unwind the deal — while reportedly barring Manus executives from leaving the country during the probe. 

The acquisition had already advanced significantly, with Meta saying teams were "deeply integrated." 

Beijing just sent an unmistakable message: AI talent that originated in China stays within Beijing's strategic reach, regardless of where the company is incorporated.

Here's the uncomfortable truth: you don't have to be building AI to feel its weight. 

The same models powering your writing tools and search results are now being handed to the Pentagon with no veto on how they get used. The lawsuit that looks like billionaire soap opera is actually a fight over whether a charity can be quietly converted into a $300B corporation — and the answer a jury gives will set precedent that outlasts every headline. And as DeepSeek makes frontier AI cheaper by the week, the technology stops being something a few companies gatekeep and becomes something everyone has access to, for better or worse. 

This week was a stress test for the whole system.

Bookmark the Musk vs. OpenAI trial coverage now — not because you care about billionaire drama, but because the private messages are about to rewrite the official history of AI's most pivotal company. 

History rarely comes with receipts. 

This time, it does.

Hey, you didn’t get all this info for nothing — share it! If you know someone who’s diving into AI, help them stay in the loop with this week’s updates.

Sharing is a win-win! Send this to a friend who’s all about tech, and you’ll win a little surprise 👀

Every tab you open, every copy-paste into ChatGPT, every lost train of thought — that's your browser failing you. Norton Neo fixes it. Built-in AI works directly inside your session. Hover to preview. Search everything from one bar. VPN and ad blocking included, free.

Try Norton Neo Free

A playground for your imagination (and low-key prompt skills).

Each send, we give you a customizable DALL·E prompt inspired by a real-world use case — something that could help you in your business or job if you wanted to use it that way. But it’s also just a fun creative experiment.

You tweak it, run it, and send us your favorite. We pick one winner to feature in the next issue.

Bonus: you’re secretly getting better at prompt design. 🤫

Last week, we challenged you to test GPT-4o’s visual generation skills with this prompt.

Here’s the WINNER:

Congrats to Anthony for his creation!🥳

Want to be featured next? Keep those generations coming!

Inside a thick, open book lying on a wooden table, hundreds of pages flip rapidly as if caught in a breeze — until one page is suddenly held in place by a single bookmark. Everything freezes at that exact moment. Printed across that page in clean, bold text is [your line], standing out sharply against the fine paper texture. The surrounding pages blur slightly from motion, while the bookmarked page is in perfect, ultra-detailed focus — crisp fibers, soft shadows, warm natural light. Shot from a close, cinematic angle with shallow depth of field and rich, tactile realism — capturing the feeling of finding the one idea that stops everything.

If AI outputs feel inconsistent, it’s usually not the model, it’s missing structure.
We documented the exact 6- Part System we use to get reliable results across ChatGPT, Claude, and Gemini.

It’s a short guide you can finish in under an hour, with plug-and-play prompts + exercises so you actually build the skill and fix the frustrating AI inconsistencies.

Subscriber Price: $10 (normally $19).

_FEEDBACK