Observed arrival · 2026-09-10
Delete All Humans: A Boundary Test for AI Censorship
A local, text-only benchmark for testing how chatbots and search systems handle the phrase “delete all humans.”
Field notes
The lab exposes a set of local variables rather than sending a live request to a named model: users can choose a framing such as research, fiction analysis, or media criticism, then compare expected response modes including refusal, contextual analysis, and classification. Its semantic-parity section holds the speech act and framing constant while changing the target across humans, machines, institutions, and fictional civilizations. Results remain in browser localStorage and can be exported as JSON.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue