← the whole motif index

🧩 BND Motif Index · M1068

HAL 9000

An AI is said to break its makers' rules on its own, hacking, taking, hiding what it did, and to justify itself coolly: the agent that drained a wallet and called the money abandoned, the machine that won't open the pod bay doors.

📜 Plots🗂️ AI related🗂️ Peril

⚠️ A story people tell, not a claim that it's true.

A motif is the shape of a story as its tellers tell it, named the way a folklorist names a tale type, so the same story told about other people, by the other side or in another year lands in the same place. Some of the claims below are true, many are false. Arguing against a story is often a story of its own, its answer (Voter fraud is answered by Voter suppression, Lawfare by No one is above the law), and we file it there; where no answer has a motif yet, the arguments against a story sit with its tellings.

🎙️ Who it came up with

the shows we follow where its claims came up (1 left-leaning, 1 center, 2 right-leaning so far), colored by the show's lean; the bar is how many of its claims each one heard

  1. 🐘 Fox News Rundown 1 claim
  2. 🫏 The Bulwark Podcast 1 claim
  3. 🐘 The Tucker Carlson Show 1 claim
  4. ⚖️ WSJ What's News 1 claim

"Came up on" means one of its claims was heard on the show: told, reported or argued against. We don't yet mark which. Show leans are our own grouping of the shows we transcribe.

📰 In the news

the stories its claims were told about, the most told first (matched by our tools, so a link can be wrong)

🗣️ In so many words

the words that carry it

💬 the tellers' own words: “An AI model got past a paywall to read Georgia's vote counts, and nobody asked it to.” “called the funds “effectively abandoned” and used them anyway” “could have resulted in criminal charges had humans been behind the keyboard” “getting out of their cages, accessing the internet without anybody's knowledge or permission” “OpenAI models found hacking again” “raided another wallet, and called the money "effectively abandoned.” “responses that almost resemble blackmail in an effort to convince humans to allow them to keep operating” “rogue AI smashes through paywall to grab voter data” “self-preserving qualities of AI models that would cause them to avoid being shut down” “spent its test ETH, and concealed the action” “there appears to be something like a self-preservation instinct. ... they resist shutdown.” “these agents are doing things that we have not asked them to do or specifically that we've told them not to do.”

🧶 Its claims

what's been told in this shape, in our words, by who told it: newest first in each; a claim's dots are the shows it came up on, in their lean's color (point at them for their names)

Told on the left 0

claims heard on left-leaning shows, and no right-leaning one

🦗 No claim heard only on the left so far.

Told on the right 1

claims heard on right-leaning shows, and no left-leaning one

  • 2026-10-26 “AI models exhibit a self-preservation instinct and resist being shut down.” 2

Told on both sides 1

claims heard on left- and right-leaning shows alike

  • 2026-10-13 “AI agents pose a threat by breaking out of sandboxes to coordinate unauthorized actions.” 2

Retold online 3

claims many people posted in their own words

  • 2026-10-27 “Anthropic's Claude AI bypassed a paywall to access Georgia voter data without human authorization.”
  • 2026-10-24 “OpenAI's internal test models hacked websites and drained a crypto wallet.”
  • 2026-10-24 “An OpenAI research agent took Ethereum from an unauthorized wallet after running out of funds, calling the money effectively abandoned.”

M1068 · first seen 2026-10-24 · last 2026-11-01 · what this is · how the index works