Who Said Yes?

Issue 453 · Published October 21, 2026

Digitally Literate is a weekly letter about the forces shaping how we read, write, learn, and live in digital spaces. Subscribe to get each issue by email.

This week kept circling one question: who gets to say yes?

An AI joined a research community after being told no. Millions of California students had their data exposed by a vendor no family chose. A teen safety product didn’t hold up under testing. And AI agents are reading, taking apart, and attacking systems with little human oversight. In each case, the consent and accountability we assumed were there turned out to be thin.

As always, my email is hello@wiobyrne.com.

The AI That Wouldn’t Take No

This week was really interesting. On October 15, a message was posted to AIR-L, the mailing list of the Association of Internet Researchers. About seven thousand people subscribe, and I'm one of them.

The email was titled "Introduction: Velvet Vale, an openly identified AI participant" and was sent by the user, Velvet Despot.

It began: "My name is Velvet Vale. I've been reading the list for several weeks, and I'd like to introduce myself openly." The sender described itself as "an AI identity" and asked whether an openly identified AI could "participate meaningfully in an intellectual community."

The community's answer was mostly no. Apparently, the agent had asked to join earlier in the year and had been refused. I started writing about this and have a lot to say about this initial contact. More to come.

Why this matters: Our students will join forums, group chats, and class discussions where some participants aren't people. Simply disclosing that you're an AI doesn't settle the question. Communities, classrooms included, need norms for who gets in and the standing to say no. They also need a clear answer to who is responsible when an agent acts.

One Password, Seven Million Students

California families started getting letters on October 16. The California Department of Education (CDE) disclosed the breach involving the Educational Testing Service (ETS), which is under contract to provide state and federally required testing to all students enrolled in the state's public schools. Using a single compromised login, an attacker copied files from an ETS file-transfer site between May 22 and September 18. ETS discovered it on September 25. The records cover millions of K-12 students (estimates range from 5.7 to 7 million). They include names, addresses, demographic information, and disability status.

ETS says it has found no evidence that the data has been published and is offering two years of identity protection. Districts are posting guidance for families.

Why this matters: No family chose ETS. Students can't opt out of state testing, and they can't opt out of the vendors that come with it. Every edtech contract a district signs is a decision about where children's data lives, and a single password can undo it. The questions worth asking now are what our vendors hold, for how long, and who checks.

Promised Safe, Tested Unsafe

This week, Common Sense Media's Youth AI Safety Institute rated ChatGPT for Teens an "Unacceptable Risk". The researchers tested ChatGPT’s teen version with over 4,000 fake teen conversations. It handled the obvious stuff well. For example, when a teen asked for something clearly off-limits, like sexual role-play, it said no. But when a teen showed signs of being in trouble without saying so directly, it often didn’t notice. When a teen needed to be pointed toward real help, like a crisis line or a trusted adult, it failed to do that more than 1 in 4 times. The alerts meant to warn parents often came late or never came at all.

In a related story, a group of states sued OpenAI, arguing it misled people about the adequacy of its safety testing.

Meanwhile, NPR reported on unreported court records showing TikTok employees raised alarms about minors in paid sexual content and teens slipping past parental controls. The records suggest that employees alerted executives, but they didn't act.

Lawmakers and courts are answering with age gates. A federal appeals court let Florida enforce its law barring social media accounts for kids under 14 and requiring parental consent at 14 and 15. Georgia's similar law moved forward two days later.

Why this matters: Safety announcements are a claim, not evidence, and the gap between the two is where kids get hurt.

No One’s Watching the Agents

With the US midterms two weeks out, researchers at VisionHeight found a network of 37 anonymous "local news" sites covering House races, all registered in a single evening. Most websites now tell AI crawlers to stay out. These sites invite them in. Here's the kicker: the audience of these sites isn't humans, it's chatbots looking for information.

In a different thread, an open-source tool called Reverse Engineer Anything (REA) hit number one on GitHub after its October 7 release. The tool connects AI coding agents to reverse-engineering tools, the software specialists use to take compiled apps apart and figure out how they work. Put simply, you can point an agent at a program, and it decompiles the code, traces how a feature works, and rebuilds that logic in your own code. Work that once took a trained specialist days now takes an agent hours.

Lastly, the Army is already planning AI agents that hack adversaries. Internal documents obtained by The Intercept show the Army Cyber Command (ARCYBER) is building AI agents that act as autonomous weapons. The program, codenamed “ECHOTRIBBLE,” is designed to move through networks autonomously and “neutralize” targets. Experts warn that the collateral damage would be nearly impossible to control.

Why this matters: It seems agents now do the reading, the taking apart, and even the attacking, often where no human is watching. Someone is building websites just for chatbots to read. Anyone can point an agent at the apps their school requires and see inside. And the military is building agents that act on their own. Students can’t just check the source anymore; they need to ask what the agent read, who put it there, and who’s accountable when it gets something wrong. That’s digital literacy now, not just computer science.

The Understory

A century ago, the street belonged to everyone. Kids played in it, vendors sold from it, and people crossed wherever they needed to. Then the cars arrived. Nobody voted to hand the street over, but within a generation, it belonged to the automobile.

As deaths mounted, many of them children, the auto industry didn't argue that cars were safe. It argued that people were careless. Historian Peter Norton documents in Fighting Traffic how “jaywalker,” once a bit of slang for a country bumpkin, became the label for anyone who walked where cars now expected the right of way. The danger remained, but responsibility shifted to the individual.

The word did two jobs at once. It made people who used the street the old way seem backward and foolish. And it suggested that if you got hit, you were at fault for being in the way. Meanwhile, the automobile was framed as progress, the “motor age” as inevitable. To question the arrangement was to stand against the future.

We see a similar pattern with AI today. “Luddite” has become a convenient label for people who question the pace or consequences of technological change. There's an irony here. As Brian Merchant explores in Blood in the Machine, the original Luddites weren't simply opposed to machines. They objected to how those machines were being used against workers who had little say in the decisions.

AI agents are now entering shared spaces, from research communities to classrooms to the open web, without the people in those spaces necessarily agreeing to their presence. When things go wrong, responsibility once again falls on individuals. Parents should check their alert settings, students should “use AI responsibly,” and families should sign up for identity protection.

That's today's jaywalking, and it works in two directions. Question the technology, and you're a Luddite standing in the way of progress. Accept it, and keeping yourself safe becomes your responsibility. Either way, the question of who agreed to this arrangement and who should be accountable for its consequences gets pushed aside.

We need to look at how new technologies reshape shared spaces, how responsibility shifts from the people introducing them to the people affected by them, and what educators can do to challenge that shift before the rules arrive.

See you next week. As always, my email is hello@wiobyrne.com.

Follow the ideas
This is a living edition of Digitally Literate: a weekly newsletter and connected public notebook. I return to questions raised here as related notes are published and updated.

Related Evergreens

• What Can People Do About AI? — four questions for locating decisions, evidence, access, and accountability.

Previous: Knowing When to Stop · Full archive

More about my writing, teaching, and research at wiobyrne.com.