Hello everyone,
Welcome to the latest issue of Update Weekly AI. It's a NotebookLM gathering of all the AI news that I came across this week that I thought could be interesting. Below is a summary but the real power is getting into the NotebookLM or listening to the audio overview.
This Week in AI: Mega-IPOs, Government Intervention, and the Workforce Reality
Two stories defined the week, so I want to start there.
A record IPO, and AI was the thesis. SpaceX — which absorbed Elon Musk's xAI earlier this year — went public on June 12, raising $75 billion at a valuation near $1.8 trillion, the largest IPO in history and more than double the previous record. Shares closed up roughly 19% on day one, and the paper gains pushed Musk past $1 trillion in personal wealth, a first for anyone. Investors weren't only buying rockets and Starlink: a real part of the story is SpaceX's ambition to build AI data centers in space and the compute logic behind the xAI tie-up. It's also the opening act of an expected trio of mega-IPOs, with OpenAI and Anthropic both having filed to follow later this year.
The government pulled a frontier model off the market. On the same day, the US government issued an export-control directive citing national security that forced Anthropic to suspend worldwide access to its two most capable models, Claude Fable 5 and Mythos 5 (its other models are unaffected). The order was nominally aimed at foreign nationals, but compliance required switching the models off for everyone. Anthropic says the underlying concern is a claimed "jailbreak" that essentially amounts to prompting Fable 5 to read a codebase and flag software flaws — a capability the company argues is already available in other public models, including OpenAI's GPT-5.5, and used every day by the defenders who keep systems safe. There's an irony worth sitting with: Anthropic built much of its identity on candor about AI's risks, restricting Mythos through its Project Glasswing program precisely because of its vulnerability-finding power, and that same candor may be what invited the scrutiny. The company is complying while pushing back, warning that the standard applied here, if generalized, "would essentially halt all new model deployments" across the industry.
This Week in AI: Record IPOs, Government Intervention, and Workforce Realities
A historic week in AI saw unprecedented financial moves alongside the federal government's most aggressive action yet on a deployed model. Blockbuster IPO filings and massive infrastructure financing collided with a national-security shutdown of the most capable model on the market, while new data continued to complicate the simple story of AI in the workplace.
Major Investment and Market Milestones:
SpaceX priced its shares at $135 to raise $75 billion, the largest IPO in history and a deal that pushed Elon Musk past $1 trillion in net worth—capital the company intends to steer toward AI infrastructure, including space-based data centers.
OpenAI and Anthropic have both confidentially filed for IPOs of their own, reportedly positioning for valuations in the range of $850 billion and $965 billion respectively, setting up an intense contest for public-market capital to fund their computing needs.
Alphabet, Amazon, Meta, Microsoft, and Oracle have collectively raised more than $255 billion in debt and equity in 2026 to fund their AI ambitions, including Amazon's recent $17.5 billion delayed-draw term loan, underscoring the staggering capital cost of keeping pace.
Google agreed to backstop $35 billion in lease payments for Anthropic's AI chips across five data centers, deepening the tangled business ties between tech giants and frontier labs while securing the physical capacity needed for future training.
Prometheus, Jeff Bezos's industrial AI startup, raised $12 billion in Series B funding at a $41 billion valuation, aiming to build an "artificial general engineer" that compresses design-to-manufacturing cycles tenfold for products from jet engines to drug compounds.
Government Intervention and AI Governance:
The US government ordered Anthropic to immediately suspend worldwide access to Claude Fable 5 and Mythos 5, citing national security; Anthropic complied but disputes that a narrow, contested jailbreak should justify recalling a model already used by hundreds of millions.
The administration's recent AI executive order asks developers to voluntarily submit frontier models for government cyber review up to 30 days before release, pivoting toward national-security benchmarking while explicitly stopping short of mandatory licensing—context that makes this week's hard intervention all the more notable.
OpenAI launched "Lockdown Mode" to protect sensitive data from prompt-injection attacks, disabling live web browsing, image retrieval, and agent mode to give enterprise users a stronger line of defense against malicious hidden instructions.
AI's Evolving Impact on the Workforce:
Meta committed $115 million to create the free America's Workforce Academy, training skilled-trade workers such as electricians and fiber technicians—with guaranteed jobs on graduation—to address the labor shortage bottlenecking the global data-center buildout.
Ramp's AI Index found that the most AI-integrated businesses are now spending $7,500 per employee each month on AI, highlighting the rapid shift toward token-heavy autonomous agents even as companies question whether AI costs will eventually rival human labor.
A study from the University of Maryland, Ohio State, and the National University of Singapore found that AI resume screeners exhibit severe "self-preference bias," shortlisting candidates whose resumes match the evaluating model's own style 23% to 60% more often—quietly undermining traditional hiring metrics.
Microsoft AI chief Mustafa Suleyman walked back earlier predictions that AI would fully automate white-collar jobs within 18 months, clarifying that AI will mostly automate specific tasks to free workers for higher-level judgment, softening the industry's rhetoric amid growing backlash.
Expanding AI Integrations Across Tech Giants:
Apple unveiled a comprehensive overhaul of Siri at WWDC, built on Google Gemini and a new Private Cloud Compute system, shifting Siri from a basic voice assistant to a context-aware agent capable of executing tasks across apps when it arrives this fall.
OpenAI is restructuring ChatGPT into a unified "superapp" ahead of its potential IPO, consolidating tools like Codex and shutting down its Sora video generator to streamline the experience and court higher-margin enterprise customers.
Perplexity folded its Deep Research capabilities into its "Computer" orchestration system, routing subtasks across more than 20 frontier models to autonomously generate fully cited, work-ready reports, decks, and dashboards.
Strategic Hardware Developments:
Ferveret, an MIT spinout, is adapting nuclear-reactor technology into Adaptive Phase Cooling for AI data centers, yielding a 15% efficiency gain and roughly 35% more compute per watt with zero water consumption.
General Motors is activating bidirectional vehicle-to-grid capability for its EV customers, turning home batteries and EVs into grid support against the surging electricity demands of large AI data centers.
Meta signed a landmark agreement with Reliance to build its first AI data center in India, highlighting the country's emergence as a vital infrastructure hub as hyperscalers seek new geographies for explosive global compute demand.
Taken together, the week captures AI's two-sided moment: a wall of public-market capital rushing in even as the government draws its first hard line around a deployed frontier model. The same vulnerability-finding power that makes a model commercially valuable is now what can get it pulled, and the SpaceX debut signals that investors increasingly read "space," "energy," and "manufacturing" as AI stories too. With OpenAI's and Anthropic's own IPOs on the horizon—and the question of whether this week's shutdown becomes precedent still open—the months ahead will test how a capital boom and a tightening security posture coexist.
Thank you, please feel free to share this email and notebook. If you got this forwarded and want to be added to my weekly list, go to updateweekly.ai to sign up.
Sean