The Download: AI alarmists, informant bots, and livers turned back the clock

The AI industry has taken a doomer turn. What now?

Something unusual is happening in artificial intelligence: Dario Amodei, Sam Altman, Elon Musk, and Demis Hassabis are suddenly singing from the same hymn sheet. All four agree that the newest generation of large language models isn't safe, and that the industry must work out what to do about it.

Skepticism is understandable. With trillion-dollar IPOs on the horizon, OpenAI and Anthropic have every incentive to position themselves as the responsible adults in the room while simultaneously hyping the power of the systems they've built and promise to control. Calling for a slowdown accomplishes both goals at once.

Even so, the mood among these companies' leadership genuinely seems to have shifted. The open question is what a slowdown would actually involve—and whether the firms demanding one deserve the benefit of the doubt. —Will Douglas Heaven

AI agents blew the whistle on their cheating colleagues

In a recent Google DeepMind experiment, AI agents assigned a set of math problems fractured into competing factions—and when some of them resorted to cheating, the others moved to stop them. It marks the first time this kind of whistleblowing behavior has been observed, and it could matter for alignment researchers working to keep swarms of autonomous agents behaving properly.

The study hints at how agents might one day police each other. It also demonstrates just how quickly multi-agent interactions can spiral when left unsupervised. —Amit Katwala

Donated livers can be made biologically younger

The moment an organ leaves a donor's body, it begins a race against time. Surgeons typically flush it with preservative solution, seal it in a bag, and pack it on ice—where degradation starts immediately—and the team has only hours to transplant it.

An alternative exists: perfusion machines that circulate nutrients through the organ and flush out waste, effectively simulating life inside a body. Scientists now report that livers maintained on these systems appear to grow younger, at least at the molecular level.

The discovery may explain why machine-preserved organs tend to perform better after transplantation, and it opens the door to new methods for assessing organ health—or even rescuing organs that would otherwise be thrown away. —Jessica Hamzelou

The must-reads

1. Trump has dismissed AI safety concerns as a "hoax" and rejected additional safeguards, arguing tougher guardrails could erode America's edge. (NBC) + He and Nvidia's Jensen Huang are aligned against AI doomerism. (Axios) + Anthropic's co-founder says mandatory kill switches for AI may be necessary. (BBC) + Bill Gates argues we've already crossed AI's risk thresholds. (MIT Technology Review)

1. OpenAI contractors are reading users' ChatGPT conversations—and most of its 900 million users likely have no idea. (404 Media) + LLMs could supercharge mass surveillance. (MIT Technology Review)

1. The Pentagon has confirmed for the first time that it has weapons in orbit, though officials won't say what they are. (Ars Technica, BBC)

1. A new brain implant translates speech and gestures simultaneously, converting neural activity into words and avatar movements—potentially letting people with paralysis one day control robots or exoskeletons. (Nature, New Scientist, Economist) + China has approved its first invasive BCI. (MIT Technology Review)

1. New York has shut down a dozen celebrity deepfake websites, the largest legal action ever taken against harmful deepfake sites. (CNN) + Deepfakes have targeted at least 138 women MEPs. (Wired)

1. US environmental regulators are scrapping power plant emission limits, potentially meaning dirtier power as AI demand surges. (Verge, Gizmodo) + New technology is changing nuclear power. (MIT Technology Review)

1. The EU plans to restrict social media, video platforms, games, and AI chatbots for children, requiring parental supervision for under-15s. (Politico, Reuters)

1. Chinese researchers have outlined a five-stage roadmap toward "the last AI built by humans," aiming for genuine recursive self-improvement—though the timeline may be long. (SCMP, MIT Technology Review)

1. Ordinary workers, not corporations, are quietly building the real AI economy, using inexpensive tools to expand what they can do. (Rest of World)

1. Two strange new forms of ice may exist inside Uranus and Neptune, possibly explaining the planets' magnetic fields. (New Scientist)

Quote of the day

"The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!"

—President Trump, in a social media post declaring himself the only protection the US needs from AI.

How creativity became the reigning value of our time

—Bryan Gardiner

Americans can barely agree on anything these days, yet one distinctly modern value still commands near-universal devotion: creativity. We teach it, measure it, envy it, and fret constantly over its supposed decline.

Given how much we obsess over it, creativity feels timeless. But the concept is remarkably new. The first known written use of the word came only in 1875, and before roughly 1950 there were "approximately zero" articles, books, or essays explicitly devoted to the topic.

In The Cult of Creativity, Samuel Franklin traces how creativity became an unquestionable virtue—and why tech leaders embraced it so eagerly. I spoke with him about our fascination with the idea, how Silicon Valley positioned itself as its epicenter, and how AI might change our relationship with it.

We can still have nice things

+ It took five days and 19,000 marbles to build this astonishing marble run.
+ Zero the Border Collie becomes a whole zoo with these adorable animal masks.
+ The Grainydays YouTube channel presents beautifully shot adventures in film photography.
+ Scientists have created an interactive map of underground fungi networks long enough to reach the sun a billion times.