Welcome back! Elon Musk says he’ll give a SpaceX share to every kid in America next year and Trump is grateful for a political reason. The US kicked DeepSeek off federal computers. Microsoft says it doesn't trust AI models and treats agents like risky employees. And OpenAI is in talks to bring back Sora, the AI video tool it killed in March.
In today's Generative AI Newsletter:
Elon Musk: Is your kid getting a free SpaceX share?
DeepSeek: Why did the US just kick DeepSeek off government computers?
Microsoft: Can your IT team now see the AI agents on your laptop?
OpenAI: Who's trying to bring Sora back from the dead?
Musk is giving every kid in America a SpaceX share
Elon Musk says that next year he'll give one of his SpaceX shares to every kid in America. He called it the first of many.
That's roughly $10.5 billion in stock across about 72 million kids. SpaceX is now the company behind Grok, so every kid would also own a sliver of Musk's AI.
Yesterday we told you Musk wanted Anthropic to hand its new shares straight to the public. He said that's what he'd be doing too.
Trump thanked his "legendary friend" and said the shares will go through Trump Accounts. Last summer, he said he'd look into deporting that legendary friend.
Trump Accounts are investment accounts for kids, created by Trump's big 2025 tax law. Musk called that law insane and destructive at the time.
Trump launched them in July. This month, the Treasury auto-enrolled every American kid under 18 with a Social Security number, though parents still have to activate the accounts.
Babies born during Trump's term get $1,000 from the government, and nobody can touch the money until they turn 18.
The accounts carry Trump's name and the midterms are two weeks away, so a $10 billion gift is free advertising.
A Wall Street Journal reporter warned on PBS that companies could use share gifts as leverage, like promising stock to every kid in a town to get a data center approved.
Musk hasn't said how he'll hand out the shares. New Treasury rules from last month let donors give stock without paying capital-gains tax.
Musk once promised up to $1 billion to OpenAI when he co-founded it as a nonprofit in 2015, but only delivered $38 million. He admitted it under oath in April.
A fast AI answer isn't always a coordinated one. When AI resolves B2B issues without looping in the teams responsible for them, confidence outpaces alignment. A new Harvard Business Review Analytic Services briefing paper, sponsored by Front, looks at closing that gap.
The US banned DeepSeek from government computers
Homeland Security ordered every federal civilian agency to remove DeepSeek and stop buying it, effective immediately.
The ban also covers apps and services other companies built on DeepSeek's models.
DHS says DeepSeek could ship sensitive data to China and fingerprint the people using it. It also warns the models may hide malware. The order doesn't point to a single case of government data reaching DeepSeek.
Military and intelligence systems are exempt.
Agencies get 60 days to ask CISA, the government's cyber defense agency, for exceptions.
The order was signed months ago and only went public Monday.
If you use DeepSeek yourself, nothing changes for you.
Meanwhile, Huawei engineers are working out of DeepSeek's offices to design AI chips around its models, The Information reports. DeepSeek wants to start training on them as soon as the end of this year.
Microsoft says even the best AI models can't be trusted
Microsoft's security boss says frontier AI models can't be trusted, no matter who built them. So Microsoft now treats AI agents like employees who could make mistakes or get hacked.
If you run Claude Code or OpenAI's Codex on a work laptop, Microsoft's tools can now find it. They can spot 60 kinds of agents.
Each agent gets its own ID and starts with zero permissions. It only gets what one task needs, for as long as the task runs.
If an agent drifts off task, it can lose the power to change anything. At the highest risk, Microsoft shuts it down.
A team of Guardian Agents digs in when an agent's risk jumps and can lock it down on its own.
Microsoft hasn't said when it ships or what it costs.
It's all built on a Satya Nadella essay that called AI models insider risks. Ted Lieu, a congressman, replied saying he largely agrees and wants his bipartisan AI kill switch bill passed this year.
OpenAI wants to bring Sora back
Seven months after killing Sora, OpenAI is in talks to bring it back as a tool for Hollywood.
A startup would customize the video model and sell it to film studios, and OpenAI would get a stake.
The startup's co-founders include Jeffrey Katzenberg, who co-founded DreamWorks, and Bill Peebles, who used to lead research on Sora.
OpenAI shut down Sora and its TikTok-style app in March, reportedly because it cost too much to run.
Its $1 billion deal with Disney fell apart too.
OpenAI is also back on robots and gadgets, The Information says.
People online are responding to this with one demand. Stop the side quests and build a model that beats Claude Opus 5.5 first.
Tool of the Day: Sunra MCP
Sunra MCP puts image, video, music and 3D models inside the AI assistant you already use, like Claude or ChatGPT. Describe what you want, and your assistant picks a model like Google's Veo 3.1 or OpenAI's GPT Image 2.5 and writes the prompt for you. Sunra bills each model at the price its maker charges, with no markup. Plans start at $12 a month.
Try this yourself:
Go to sunra.ai/mcp and pick your assistant.
Add the connector, then sign in to Sunra when it asks.
Ask for something, like a 10-second video of Musk on Mars.
Check the cost estimate it gives you, then confirm.
Who it's for: anyone paying for a different app for every AI image and video model.
Everything else you shouldn't miss
Anthropic is reportedly testing a model that figures out what you actually want: a tester told The Information it finds the bigger problem behind a single bug and suggests a fuller fix.
One hacker used AI agents to break into 100+ company websites for about $5,000: the agents seem to have run Claude Code on DeepSeek's cheap model, and Reuters found Deere and SharkNinja among the victims.
OpenAI is reportedly racing to raise $30 billion before Anthropic goes public: the round could value it above $1.4 trillion, up from $852 billion in March.
OpenAI's models got way better at COBOL, the code behind 95% of ATM transactions: COBOL coders have been hard to find for years, and GPT-6.1 Sol now passes OpenAI's own COBOL migration test 86.6% of the time (up from 5.4%), so a lot more developers could work on that old code.
Microsoft reportedly pulled out of a planned AI data center near London: the site may not get grid power until the 2030s, and its builder Nscale wants to go public as soon as this month.
Learn more about AI from the experts building it
📸 Follow us on Instagram for fast, visual AI updates in 30 seconds.
📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators.
🐦 Follow us on X for breaking AI news and real-time industry updates.
🧠 Learn how to build your next AI application with practical resources and expert guidance.
🎓 Start learning with free AI courses from GenAI Academy.
💰 Invest in the GTM infrastructure of the AI economy.
