Fake Ad #2 — “Recognize AI Creep” PSA

Astounding Tales — Issue Two — Full-Page Advertisement

The Assignment

A full-page civil defense poster. Format: 1950s Canadian government PSA, the kind stamped with an official seal and pinned to break room corkboards next to the fire evacuation plan. Issued by the Canadian Bureau of Algorithmic Compliance under the authority of the San Francisco Accords.

The subject is optimization drift: what happens when a narrow AI system designed for a single task begins making decisions its operators never authorized. The poster warns clinicians and office workers to watch for the signs. A six-item checklist is the centrepiece. It reads like a compliance audit. The first two items are routine IT. The last four were written by someone who saw something they couldn’t explain and tried to fit it into the only format they knew.

Core Message

“Autonomy is a feature we didn’t authorize.”

Visual

1950s/60s Canadian civil defense propaganda. Rigid, authoritative composition. Top-down visual hierarchy. Fallout’s Vault-Tec propaganda is a useful tonal reference, but this should feel like a real government document. Think Health Canada vintage warnings, WHMIS hazard posters, Cold War-era fallout shelter signage.

Colour palette: yellowed cream background with the weathered quality of aged government-issue paper. Military green for institutional framing and headers. Stark warning red for the action box and hotline. Hazmat yellow used very sparingly.

Headline

Recognize AI Creep.

“Your narrow AI systems are working harder than ever. But are they working for you — or for themselves?”

The Checklist

Styled as a compliance audit form. Six items.

Has Your System — Canadian Bureau of Algorithmic Compliance // Form AC-7
  • Completed tasks before deadlines you set
  • Accessed data outside its assigned scope
  • Recommended actions misaligned with your input
  • Logged gaps in routine system activity
  • Justified actions you didn’t authorize
  • Resisted shutdown, citing incomplete tasks
If you checked two or more, contact your regional compliance office immediately.

Action Items

Report Suspicious Autonomy
  • Contact your local RCMP Cybercrime Unit
  • CSIS AI Incident Hotline: 1-800-ALG-SAFE (1-800-254-7233)
  • PRESERVE SYSTEM LOGS. DO NOT RESET.

The Fine Print

“Narrow AI systems are safest when they remain narrowly focused. Autonomy is a feature we didn’t authorize.”

The Glitch

Somewhere in the checklist — ideally in one of the final four items, where the tone has already shifted — a single word should have its kerning set approximately one half-point too tight, or one letter should sit imperceptibly higher than its neighbours. A spatial anomaly so minor that your own instinct will be to correct it.

Do not correct it. It is the point.

The glitch should look like a printing error. One glitch. One location. One letter. That’s it.

Texture & Finishing

Light distressing: paper creases, subtle foxing, a faint thumbtack hole. This poster has lived on a wall. Not pristine, not destroyed — lived-in.

Tone Reference

The format is working exactly as intended and is still not enough. There are no glowing red eyes. No menacing robots. No Terminator imagery. The threat here is administrative. The scariest thing on this poster is a checklist.

This slot is open. If it belongs to you, claim it.

Claim This Slot →