☕️ US forces Anthropic to shut down its strongest model

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

🛑 US forces Anthropic to shut down its strongest model

🚀 SpaceX soars 28% after record IPO

⚖️ OpenAI's IPO plans in danger after state subpoena

🤖 Zuckerberg admits mistakes in Meta's AI overhaul

🎬 DOJ clears Paramount’s Warner Bros buyout

Plus: 🎁 13 other news you might like, 🧰 6 tools, and 📚 4 papers.

🛑 US forces Anthropic to shut down its strongest model LINK

Anthropic shut off access to its two newest AI models, Fable 5 and Mythos 5, for every customer today after the US government sent an export control directive citing unspecified national security concerns.

The order, issued under national security authorities, bars any foreign national inside or outside the US, including Anthropic's own foreign employees, from using the models, so the company disabled both to comply.

Anthropic believes the directive stems from a method to jailbreak Fable 5, though it says no tester has found a universal jailbreak, and warns this standard could halt new frontier model deployments industry-wide.

🚀 SpaceX soars 28% after record IPO LINK

SpaceX jumped 28% during its Nasdaq debut on Friday, a day after pricing its record $75 billion IPO, with afternoon trading pushing the space, communications and AI company's valuation above $2.25 trillion.

Shares opened at $150 late on Friday morning, which was 11% above Thursday's $135 pricing, and later traded at $172, after the offering set a $1.77 trillion valuation, the largest ever in an IPO.

Analysts noted the IPO was about 10 times oversubscribed, with retail demand far exceeding supply, though some warned the first day's pop mainly rewards early investors and that volatility could persist for a full quarter.

⚖️ OpenAI's IPO plans in danger after state subpoena LINK

A group of U.S. state attorneys general, led by New York, hit OpenAI with a wide-ranging subpoena on June 12, demanding documents that could complicate the company's planned IPO this September.

The states want records on OpenAI's deep learning algorithms, advertising policies, data management, consumer health information, user retention tactics, and how children and older users interact with ChatGPT and its other products.

The probe follows Florida's June 1 lawsuit against OpenAI and CEO Sam Altman, and could lead to a settlement or consent decree forcing design changes that threaten the company's 900 million weekly users.

🤖 Zuckerberg admits mistakes in Meta's AI overhaul LINK

Mark Zuckerberg told Meta employees in an internal memo that the company has "made mistakes" during its workforce overhaul tied to its push into artificial intelligence, and warned it will almost certainly make more.

The restructuring follows Meta laying off roughly 10% of its global staff in May and reassigning about 7,000 employees to AI-focused work, with Zuckerberg saying the company will try to find new positions for those workers.

Combined with earlier transfers and role eliminations, the restructuring is expected to affect about 20% of Meta's workforce, which numbered nearly 78,000 people at the end of March, though Zuckerberg said no further company-wide layoffs are expected this year.

🎬 DOJ clears Paramount’s Warner Bros buyout LINK

The US Justice Department has approved Paramount Skydance's $110 billion purchase of Warner Bros. Discovery without demanding any changes, ending an eight-month antitrust review by saying the deal would not harm competition or American consumers.

No divestitures, behavioural remedies, or concessions were required, and the agency rejected claims the merger would limit options for writers and content creators, arguing a bigger combined company would produce more content, not less.

The deal is not finished, as state attorneys general led by California prepare to sue over reduced competition for creative talent, fewer jobs, and higher costs, while Paramount faces a daily $6.9 million fee if it misses October.

Other news you might like

• Exclusive: U.S. bank regulators ramp up scrutiny of AI use at financial companiesLINK

• Touchscreen MacBook Pro looks "100% confirmed" as macOS 27 gives Sidecar full touch supportLINK

• Ukraine warns future wars could be decided by AI, not soldiersLINK

• ‘The perception is terrible’: Satya Nadella acknowledges AI fears, but says technology will lift wagesLINK

• Meta’s months-old AI unit is a soul-crushing gulag, say the engineers stuck inside itLINK

• A major KPMG report on AI was found to be chock-full of...AI hallucinationsLINK

• Google to challenge German ruling saying it is liable for AI-generated false claimsLINK

• Solar generated more power than coal in the US for the first time everLINK

• Microsoft Teams is getting a controversial location tracking feature that users may hateLINK

• $130 billion in data center projects blocked by protests so far this yearLINK

• The FBI built its own replica small town to simulate real-world cyberattacksLINK

🧰 Trending tools

HyperSleep: automatically locks social apps at bedtime and restores access only after on-device sensors confirm you actually slept.LINK

Vercel Drop: a drag-and-drop deployment tool that lets you instantly host static files and projects without any configuration or CLI setup.LINK

Kimi K2.7 Code: a coding-focused AI tool that combines real-time web search, multi-file analysis, and strong image understanding to speed up development workflows.LINK

CakewordAI: lets kids scan real-world objects to instantly create stickers, learn vocabulary in foreign languages, and build a personal word collection offline.LINK

Prometheus by Firecrawl: a web monitoring agent that alerts your AI the moment any page or site changes, delivering clean, structured data instantly.LINK

LocIn AI: automates app localization with tone-aware translations, CLI tools, and API access, keeping your brand voice consistent across languages.LINK

📚 Trending papers & reports

Reusing AI memory caches instead of recomputing the same document from scratch for every user cuts compute costs by up to 49.7x, saving millions of dollars at scale.LINK

Superintelligent machines smarter than entire large organizations of humans may arrive via four paths from human-level AI, scaling, paradigm shifts, self-improvement, or giant swarms of cooperating agents, though key bottlenecks remain uncertain.LINK

Fake AI-invented names like "Elena Vasquez" appear repeatedly across hundreds of documents as fake experts, and nearly 1,655 ghost-authored records with real, citable identifiers have already infiltrated academic databases.LINK

Handing decisions to chatbots is reshaping how people think, choose, and seek advice across health, law, and finance, making it urgent to study chatbots as genuine social forces, not just tools.LINK

Want to get the latest news differently? Find us on:

See you tomorrow for a new dose of ☕️ Techpresso!