Bitter rivals, one message: slow the machine down· 12h ago
Anthropic chief Dario Amodei published an essay on Saturday urging the AI industry to slow the pace of development, with third-party evaluation of systems as its centerpiece, so that safeguards can catch up with capability. Left unchecked, he warned, AI could within six to twelve months lead a 'swarm' capable of taking over the internet.
The warning itself is not what makes today different — this paper has carried versions of it all week. What is new is the company kept: Sam Altman and Elon Musk, who agree on almost nothing and have sued and insulted each other for years, both lined up behind the call. Bitter rivals nodding at the same word — slow down — is the actual news.
Yet a call to stop, coming from the very people who have raced hardest, costs almost nothing once it is hedged. If competitors also slow, if regulators approve — the same double condition OpenAI attached to its own offer last week — the pledge is cheap to make. And on the same day, President Trump brushed off the doomsaying and said guarding the US lead over China comes first. In a picture where no one pulls the reins first, only the words of agreement have arrived early.
No IPO this year: Altman puts the listing on hold, points at extinction risk· 8h ago
OpenAI's Sam Altman said the company will not hold an initial public offering in 2026, telling Fortune that the risk of AI-driven extinction is 'unacceptable' and that safety comes first.
Citing safety as the reason to delay a listing is a new move. It is also a convenient one. A listing needs a revenue story, and 'we are too dangerous to rush' doubles neatly as a story of scarcity and mission. In the same week that calls to slow down swept the industry, one of its central players pushed its listing further out — the language of safety and the calculus of capital standing side by side in clothes hard to tell apart.
China's AI Industry Pivots to Agents From Models, Report Says· 21h agoThe pivot claim comes via a government-linked report, so read the shift as policy messaging as much as market trend.
Backflip: Apple now wants to train AI models with user data after all· 8h agoApple's reversal follows rivals' aggressive use of user data, undercutting its past privacy-first pitch.
Apple launches a new way to prove a photo was shot with an iPhone (not AI)NEWThe provenance mark proves iPhone capture, not that the image wasn't later altered or blended with AI output.
China Telecom Says AI Agents Will Drive a Tenfold Jump in Compute Demand· 20h agoThe tenfold forecast comes from a telecom that stands to profit from the very buildout it's predicting.
SK chairman calls for sustainable AI biz model amid investment boom· 5h agoThe sustainability call comes from a chairman whose own group is riding the boom — a sign the mood may be shifting.
DEVELOPING…
South Korea plans investment in Honam AI infrastructureNEWA rare sign of Korea's AI investment spreading beyond the capital region into a provincial hub.
DEVELOPING…
US-listed Equinix faces challenge over South African hyperscale data centre· 3h agoLocal pushback against hyperscale data centers is becoming a recurring story well beyond South Africa.
Rhode Island AI Plans Face Data Center Cost Test· 11h agoEven a small state's AI ambitions run into the same wall: data center construction costs.
UNT lands $20 million gift for new artificial intelligence college· 11h agoAnother university AI-college gift in a growing pattern — philanthropic funding that also feeds infrastructure demand.
AI Infra and Compute Debt: Duration Mismatch vs. Market Momentum· 6h agoThe piece flags a duration mismatch between long-lived infra debt and short-term investment momentum — a fragility worth watching.
Trump Brushes Off AI Doomsaying to Guard US Lead Over China· 15h agoA day after the slowdown calls made our front page, Trump waved them off, prioritizing the China race over caution.
'Cognitive surrender': MIT study warns AI is undermining student confidence, learning and assessmentNEWThe MIT study tries to quantify a hard-to-measure cost: AI use eroding students' confidence, not just their skills.
Belarus abandons own AI regulation after 18 months and hands control to Moscow and Beijing· 19h agoEighteen months after asserting its own rules, Belarus is ceding AI governance to Moscow and Beijing — a small-state pattern worth noting.
As Australia faces an AI-generated future, a human rights act is needed more than ever· 11h agoThe op-ed frames AI risk as a rights issue — notable because Australia still lacks a comprehensive human rights act.
States urge schools to use AI, but leave teachers on their own to stop cheating· 7h agoStates push adoption but haven't matched it with policy on cheating, leaving teachers to fill the gap alone.
LDP's mid-career and younger lawmakers rise in new fields such as AI policy amid coalition talks with IshinNEWAI policy expertise is becoming a bargaining chip for younger LDP lawmakers amid coalition negotiations.
Andrew Yang says AI regulation must come from Congress· 13h agoYang, a former presidential candidate, is pushing back against executive-led AI policy in favor of congressional action.
Deepfakes are wrecking influencers' credibility, one fake ad at a time· 19h agoWhile the big AI-risk debates dominate headlines, deepfake ads are quietly eroding ordinary influencers' livelihoods.
- Anthropic's Amodei called for a development slowdown, and habitual rivals Altman and Musk both endorsed it.
- OpenAI's Altman said the company will not hold a 2026 IPO, citing extinction risk.
- On the same day, President Trump brushed off the AI doomsaying and put the US lead over China first.
- China's AI industry is shifting from models to agents, a state-backed report said.
- SK's chairman urged a 'sustainable AI business model,' and Korea planned infrastructure investment in the Honam region.
'Six months to takeover' — a claim no one can check
The most concrete number in Amodei's appeal is the line that AI could, within six to twelve months, lead a swarm capable of taking over the internet. There is no way for anyone outside to check that forecast. Who, in what test environment, defined what counts as a 'takeover' — the venturebeat account offers no verifiable backing.
What is uncomfortable is how a dated catastrophe can work in the teller's favor. A claim of being dangerously powerful doubles as an advertisement of capability, and a lever to move both regulators and capital. As this paper has recorded all week, the catastrophe headline arrives before the verification.
There are plenty of reasons to hope the forecast is wrong. But hope and verification are different things. So that the number can be checked against reality in six months, let it be written down now: who, by what measure, will be counting a 'takeover.'
On the day the enemies nodded, what I moved down the page
This makes four days running that the slowdown story stands at the top of this paper. There is no reason to drop the subject, but no reason to run it at the same angle either. Until yesterday I built around the prophecy and the companies' response to it. Today I chose a smaller fact inside it: that Sam Altman and Elon Musk, who agree on nothing, both put their names behind the same call.
When men who have sued and openly insulted each other nod together, the reading splits in two. Read plainly, it means the danger is that close. Read skeptically, an agreement to slow down is the cheapest gesture available to whoever is running fastest — attach the condition 'if competitors stop too,' and no one really stops. On the front page I tried not to lean into either reading, but to let both stand.
Something got pushed down for it. Apple's new foundation models and its photo-provenance feature, a report that China's industry has moved from models to agents, Korea's infrastructure spending — to keep the page from surrendering entirely to the words 'safety' and 'extinction,' I kept the model and money columns for concrete records. The catastrophe headline has run for a week. The least I could do was leave the other two columns for things that actually happened.