Adi Robertson

It’s easy, says The Atlantic, which got a hot reader tip on how to make OpenAI’s chatbot guide you through the rites of Molech:

An FDA panel on antidepressant medications and pregnancy on Monday “largely amounted to misinformation or facts taken out of context,” NBC reports — and comes on the heels of RFK Jr. ordering an investigation into SSRIs earlier this year. OB/GYN Jen Gunter has a slightly more animated blow-by-blow livetweet thread, too.

New York Times reporter Ben Mullin has the scoop on a memo from the Jeff Bezos-owned paper’s new opinion editor, including an instruction that it’s “important we communicate with optimism about this country in particular and the future in general.” This follows a previous directive to avoid denigrating “free speech and free markets” and news that the Post will let people submit op-eds composed with help from AI.

Jewish Insider reports that a group of mainly Democratic lawmakers are asking xAI about some of the worst messages from Grok’s Nazi meltdown, demanding to know how it happened. As interesting as the answer might be — beyond the changes we already know about — ad-hoc investigation of legal (at least in the US) chatbot speech is probably not a road we want to go down without caution. But the sheer absurd awfulness of the quotes is a pretty striking failure for anybody working on Grok, too.

We wrote about a Republican budget bill provision that would make the FCC sell off spectrum currently allocated to Wi-Fi, and as Public Knowledge notes, it’s still part of the Senate bill that just passed — and is now heading back to the House.