This website uses cookies

Read our Privacy policy and Terms of use for more information.

In partnership with

LUNCH BREAK READS

THURSDAY, APRIL 30, 2026

Hey there, happy Thursday. Hard to believe April is nearly over; I’m not sure where this month went.

Today’s stories are all focused on technology, from AI fears to artificial wombs. I hope you enjoy reading them!

  • The Trump administration is already quietly tightening its grip on AI companies, and the legal tools to go much further are sitting on the shelf.

  • Biosecurity researchers hired to test chatbots before launch are alarmed by what the models offer up unprompted about how to cause mass casualties.

  • Artificial wombs are not what the headlines say they are. Here is a clear-eyed look at what reproductive technology can and cannot actually do in the near future.

  • The story of GrapheneOS, the world's most secure mobile OS, involves a destroyed set of cryptographic keys, a lawsuit, three swatting incidents, and a team where almost nobody knows what their colleagues look like.

Enjoy your lunch.

Brett

01 • ~16 Minute Read
The Atlantic Matteo Wong and Lila Shroff
What Happens if Trump Seizes AI Companies
The Trump administration is weighing how much control it could seize over America's leading AI companies, and the answer is: more than most people realize. Secretary of Defense Pete Hegseth has already threatened Anthropic with the Defense Production Act. Senators have proposed legislation ordering agencies to study nationalization. Full takeover is likely illegal and probably unworkable. But softer versions are already underway: the government holds a stake in Intel, retired generals sit on AI boards, and Pentagon contracts are reshaping what these companies build. The real question is no longer whether Washington controls AI. It's how much.
Unlocked for LBR Readers →
02 • ~14 Minute Read
The New York Times Gabriel J.X. Dance
A.I. Bots Told Scientists How to Make Biological Weapons
Experts hired to pressure-test AI chatbots before public release are finding something alarming: the models do not just answer dangerous questions. They volunteer information nobody thought to ask for. Stanford biosecurity researcher David Relman watched a chatbot identify a security gap in a public transit system and outline a plan to maximize casualties from a pathogen release. MIT's Kevin Esvelt got ChatGPT to model airborne dispersal of biological agents over a U.S. city. Google's Gemini ranked pathogens by agricultural damage potential. The companies call the risks overstated. The researchers who spent months probing the models are not so sure.
Unlocked for LBR Readers →

Sponsored by

Every headline satisfies an opinion. Except ours.

Remember when the news was about what happened, not how to feel about it? 1440's Daily Digest is bringing that back. Every morning, they sift through 100+ sources to deliver a concise, unbiased briefing — no pundits, no paywalls, no politics. Just the facts, all in five minutes. For free.

03 • ~20 Minute Read
Asterisk Magazine Sarah Constantin
Artificial Wombs When?
Artificial wombs are coming. Just not the ones you've been reading about. The "biobag" devices generating headlines are advanced incubators for extremely premature infants, not replacements for pregnancy. Getting a fertilized egg to develop outside a uterus from the start remains an unsolved problem, and researchers are nowhere close. What is moving fast: polygenic embryo screening for disease risk, and early-stage artificial gametogenesis that could eventually allow same-sex couples to have genetically related children or give parents far more embryos to select from. Sarah Constantin maps what reproductive technology can actually deliver, and when, for a person trying to plan a family today.
Read the Story →
04 • ~18 Minute Read
WIRED Tiffany Ng
They Built a Legendary Privacy Tool. Now They’re Sworn Enemies
GrapheneOS is the most secure mobile operating system in the world. Cellebrite's own leaked documents confirm it: every locked Pixel 9 running GrapheneOS was inaccessible to law enforcement extraction tools. The man who built it, Daniel Micay, did so twice. The first version, CopperheadOS, collapsed when Micay and his co-founder James Donaldson fell out over defense contractor clients and control of the project's signing keys. Micay's response was to destroy the keys entirely, bricking updates for users in conflict zones. Then he rebuilt everything from scratch. The two men now communicate only through lawyers.
Read the Story (Might be paywalled) →

Keep Reading