Sarmadi AI Digest June 18, 2026 Updated 6:55 AM CT Today Archive Topics Saved Subscribe RSS

Only 16% of Americans see AI as positive; world leaders want US AI without the US kill switch

Two-thirds of Americans say AI is advancing too quickly and only 16% expect a positive societal impact — the Pew and TC surveys land the same week SpaceX hit $2.6T and OpenAI's losses leaked. Macron and Modi pressed the G7 on the obvious corollary: countries want American AI but not the ability for Washington to turn it off, exactly as the Fable shutdown demonstrated. New details on that shutdown: Wired traced the Mythos export incident to Korea's SK Telecom, and the White House is now demanding Anthropic block every jailbreak before Fable can return — Wired explicitly says that may not be possible. Midjourney pivoted from cat images to medical AI ultrasound hardware. World-model maker Odyssey closed a $1.45B round backed by Amazon. NVIDIA disclosed self-improving robot training run by swarms of AI coding agents. And Pramaana raised $27M from Khosla to bring formal verification to AI for law, drug discovery, and tax.

8 papers 22 news 7 sources ← Latest

News

19 items

Public AI sentiment turns hard

Only 16% of Americans expect AI to have a positive societal impact; Pew finds two-thirds say it's advancing too quickly. At the G7, Macron and Modi raised the related foreign-policy concern — countries want American AI but not the kill switch Washington just demonstrated. Meanwhile, the 'slowtech' counter-trend is becoming a TC headline category.

News TechCrunch AI

Only 16 percent of Americans think AI will have a positive impact on society

TC: just 16% of Americans expect AI to have a positive societal impact — Wall Street's enthusiasm sits against deep public skepticism.

positive-impact view 16%
Why it matters
  • Lowest-recorded positive-AI sentiment in a major US survey to date.
  • Direct counter to the IPO-summer narrative — political-economy gravity now visibly tilted.
  • Makes Florida-style state AG actions and Illinois-style safety laws much easier to push.
policymarketcommunity
Source →
News TechCrunch AI

World leaders want American AI. They just don't want America to be able to turn it off.

TechCrunch on Macron and Modi raising the post-Fable concern at the G7 — US AI dependency under a kill-switch precedent.

Why it matters
  • Names the geopolitical risk the Fable shutdown actually demonstrated, not the safety risk it claimed to address.
  • Strengthens the sovereign-AI argument across France, India, Brazil, Netherlands.
  • Likely accelerates non-US export-controlled-but-redistributable open-weights deployments.
policyregulationmarket
Source →

Fable saga: SK Telecom link, 'block all jailbreaks' demand

Wired pinpoints South Korean carrier SK Telecom as the actor at the center of the Mythos export incident, refining the China narrative. The White House is now demanding Anthropic block every jailbreak before Fable 5 can return — a technically dubious requirement Wired calls out as 'may not be possible.' Stratechery and The Verge continue to dissect the standoff.

News Wired AI

The White House Wants Anthropic to Block All Jailbreaks. That May Not Be Possible

Wired: Trump administration officials told Anthropic that Fable 5 stays offline until every jailbreak is blocked — likely an unmeetable bar.

Why it matters
  • Sets a technically dubious release standard that any future frontier model would also face.
  • Implicitly favors closed, harder-to-probe deployments over auditable ones.
  • Pairs with the 'dangerous models coming no matter what' framing as the structural counter.
policyregulationsafetyproducts
Source →

World models attract real capital; Midjourney pivots into hardware

Odyssey closed a $1.45B round backed by Amazon to keep building world models for physical AI. Midjourney showed its first hardware product — a medical-AI ultrasound device, a sharp pivot from cat images. NVIDIA disclosed self-improving robot training run by swarms of AI coding agents. Kairos and PAIWorld papers add native world-model stacks designed for production physical AI.

News The Verge AI

Midjourney goes from generating cat images to full-body ultrasound scans

Midjourney CEO David Holz unveiled the company's first hardware product — a medical-AI ultrasound device, a sharp pivot from consumer image gen.

Why it matters
  • First major repositioning of a consumer-image-gen darling into regulated medical hardware.
  • If real, opens an entirely different regulatory and reimbursement path than the Hollywood-AI lane Midjourney was on.
productsimage-generationmarket
Source →

AI economy meets infrastructure: ROI, formal verification, sovereign capital

NEA partner Tiffany Luck said enterprises are 'still figuring out their AI ROI' — the most measured framing of the buyer-side reckoning yet. Pramaana raised $27M from Khosla to bring formal verification to AI for law, drug discovery, and tax. A Canadian pension giant joined the race to fund India's AI data-center boom. Anthropic became the first AI startup to join the Frontier carbon-removal coalition.

Consumer-AI hardware lands mixed: Snap dives, Google Home ships, Adam open-sources CAD

Snap's $2,195 Specs glasses unveil pushed the stock down. Google's first new smart speaker in six years ships June 25 — Gemini front and center. Adam (YC W25) launched as open-source AI CAD on HN (187 points).

Papers

2 items

World models attract real capital; Midjourney pivots into hardware

Odyssey closed a $1.45B round backed by Amazon to keep building world models for physical AI. Midjourney showed its first hardware product — a medical-AI ultrasound device, a sharp pivot from cat images. NVIDIA disclosed self-improving robot training run by swarms of AI coding agents. Kairos and PAIWorld papers add native world-model stacks designed for production physical AI.

Also today