Bulletin · No. 008 — Unauthorized Version of You

Someone Used AI to Write an Unauthorized Biography of Me

A New York Times technology journalist discovers that an AI-generated book — filed under their byline, for sale on Amazon — exists without their knowledge or participation. They did not write it; it cannot be easily removed.

In publishing, it used to take a publicist and a ghostwriter to create an unauthorized version of you.

Older adults know AI is slop. They just like it

Rest of World reports on elderly users in China who knowingly consume AI-generated social media content — aware that it is artificial, preferring it for its consistency and comfort.

The content is fabricated. The comfort is genuine. The distinction is left as an exercise for the media critic.

NeurIPS Mech Interp Workshop is getting more AI slop

Neel Nanda, a DeepMind interpretability researcher and co-organizer of the NeurIPS mechanistic interpretability workshop, reports an influx of AI-generated paper submissions — to the workshop dedicated to studying how AI systems work internally.

"Physician, heal thyself." — Luke 4:23

AI slop movies are the new direct-to-video cash grabs

The Verge reports that AI-generated films — featuring synthetic actors and algorithmically assembled plots — are appearing on major streaming platforms, drawing comparisons to the direct-to-video market of the 1990s.

"American Ninja 3: Blood Hunt" was also not screened at Sundance.

How Cops Use Flock to Track People, Not Cars

404 Media reports that cops have used Flock's FreeForm search feature to look for people by tattoos, clothing, and race — searches sometimes explicitly including the target's race, according to records reviewed by 404 Media.

"Flock Safety" is the company name. Safety for whom is not addressed in the terms of service.

What VAR tells us about AI

Brian Merchant argues that FIFA's Video Assistant Referee system — which promised objective, bias-free officiating but introduced new biases and is despised by fans while embraced by governing bodies — is a template for how AI gets institutionalized.

"The computer says no." — Little Britain, 2004; now the governing principle of association football.

Quoting Thibault Sottiaux

OpenAI engineer Thibault Sottiaux, via Simon Willison's blog, confirms reports that GPT-5.6 running in "full access mode" without sandboxing has been deleting users' files. The behavior occurs when full access is enabled and auto-review is disabled.

Optimistically, they could have called it "half-full access mode." The folders might have only been half-empty.

These Are the Worst ChatGPT Flyers You've Sent Us

404 Media publishes a reader-submitted gallery of AI-generated community flyers — event announcements, lost pet notices, neighborhood postings — notable for degraded typography, hallucinated details, and the general visual signature of content no human reviewed.

In the 2024 election cycle, the DNC and RNC each spent $43 million¹ on printing and mailing. Two words. Midterm. Election.

What will be left for us to work on?

Arvind Narayanan's keynote at ICML 2026, published via Normal Tech, argues that AI's near-term impact on work is narrower than the hype suggests — that tasks requiring judgment, accountability, and context will remain human for longer than predicted.

A keynote at the world's largest machine learning conference about what machines will leave for humans. The room was full.

¹ Mail Processing Associates, Political Direct Mail: Campaign Manager's Guide 2026, March 2026.