Someone forwarded you this? Get your own Friday Four →
Hi, we're Logan and Mara. We make a podcast figuring out AI without the ego.
Every Friday at 4pm EST we share our takes on one AI story that's probably bad news and one that's maybe good news, then pick something for you to read this weekend. Every take has a voice note, so read it or tap ▶ Play. We don't always agree.
01 · THE (PROBABLY) BAD NEWS
London scanned half a million faces at train stations. Zero arrests.
British Transport Police ran a six-month live facial recognition trial at London railway stations. It cost more than £320,000 and almost 100 hours of officer time. It produced one alert, and that one was wrong. Transport for London backed extending it, saying it would "specifically tackle violence against women and girls." The trial is now expanding to the Underground. Liberty Investigates →
MARA
The privacy risks here are huge, and the pilot looks like a complete failure. But I have to say, the fact that the policy community recognizes public transportation can be dangerous for women in a way it isn't for men is remarkable. You need it to get to school, to get to work. When that environment is unsafe, it's incredibly isolating.
Now, this technology doesn't look fit for purpose at all. It didn't work, it wasted a lot of money, it grabbed up tons of data it didn't need, and it could end up mistargeting people left and right. It feels like we're living in 1984.
The glimmer of a bright side is that women's different concerns are getting attention in policy circles.
LOGAN
This one feels like it should be good news. I'm glad money is being spent to protect women on public transportation and tackle violence against women and girls. But rolling out widespread facial recognition with absolutely no record of success feels a little like a pipe dream.
There's a well-documented 2019 study on how often this technology misidentifies women and people with darker skin. The tech has almost certainly improved since then. But this is a lot of money spent on watching people, and it doesn't seem like the most direct way to solve the problem.
If I get groped on the tube, what, they're going to track somebody down? Is the idea that Big Brother is watching going to stop it? This doesn't fill me with confidence.
02 · THE (MAYBE) GOOD NEWS
The FTC is investigating OpenAI and Anthropic.
The Federal Trade Commission has opened an investigation into OpenAI, Anthropic and other AI companies over the dangers their products may pose to consumers. It's using its consumer protection powers and preparing formal demands for information. It comes weeks after a run of incidents with AI agents acting outside their intended limits. AP →
MARA
The FTC is investigating OpenAI and Anthropic under its consumer protection powers, and I think it's great. It's finally an example of government working the way it's supposed to, in the middle of all this reckless enthusiasm for the next model at all costs. This is the government stepping in and saying, hold up, peeps. Let's take a minute and look at what's actually happening to people.
And it comes after a couple of weeks of rogue agent escapes and AI agents running cyberattacks on the Australian government. That's before you even get to the risks people face one by one, like teen suicidality and AI psychosis.
There's a lot that's amazing about AI. There's also a lot we don't understand. We need protections in place so we can get the benefits without absorbing all the risks.
LOGAN
This gives me whiplash, which is on brand for this administration. The timing is wild. On Tuesday, Trump said he'd never stifle AI and he's not into guardrails. That same day, the heads of OpenAI, Anthropic, Google, Meta and Nvidia signed a pledge to police themselves. He called it "morally binding," which is rich. A couple of days later, the FTC opens an investigation into two of those companies.
So which one is the real signal? Maybe both. Maybe "we won't slow you down" and "we'll come after you if you hurt people" can live in the same government. But I'm not holding my breath. I want to see this investigation actually produce something before I believe the stick is real. Self-policing has a lot of question marks around it.
THE TENSION
Neither of us is done with this one.
Mara: Finally, the government's saying hold up. Let's see what this is doing to people.
Logan: The same week, it promised not to. I'll believe the stick when it lands.
Forward this to the friend who'll have a different take.
03 · MARA'S WEEKEND READ PICK
Trump says top tech firms have signed accord to 'self-police' AI
NPR · September 30
I chose this one because it's the closest we've come so far to shared safety norms for AI companies. People should know what the accord actually says. And it's worth thinking hard about putting faith in self-policing, which counts on every company moving in lockstep under enormous pressure. Alex Pascal, executive director of the Berkman Klein Center, is quoted in it: "The only way to reduce the risks and harms of AI to Americans is through robust legal liability, regulation and fundamentally changing the race dynamics that are driving unsafe decision-making by frontier AI developers whose costs are borne by the American people." Pretty well said.
04 · LOGAN'S WEEKEND READ PICK
Will A.I. Make Your Brain Lazy? Here's What the New Research Actually Shows.
Juliana Castro Varón and Dylan Freedman, The New York Times · Machine Learnings column · September 29
I read a lot about cognitive offloading and cognitive surrender, and I really enjoyed this one, probably because the illustrations and the interactivity make it so digestible. It's the article you can send to the friends who ask, "Is AI going to kill us? What should I be worried about?" A good entry point into what AI does to our brains. I hadn't seen the brain's habit of conserving effort compared to how we hang on to calories. That was really interesting to me.
How we use AI to get smarter instead of lazier is the part I care about most, and I don't think the piece went far enough into that side. Still, a really good primer on a very important topic.
YOUR TURN
What are we missing?
Hit reply and tell us the part that's still nagging at you. A sentence is plenty. Rather say it out loud? The machine's on.
WRITTEN (AND SAID OUT LOUD) BY
Logan Currie & Mara Bolis
Women, work, and the messy middle of AI.