All systems going well.

Humanity uptime: 100.000% (so far)

0days since the last Going rogue incident

0%average p(doom) cited on this site by people in the field

0incidents logged here of AI agents going outside their instructions or their sandbox

Incident history

one bar per month · worst status shown

Jan 2023Today

Going well Going fine Going sideways Going rogue

Incident log

every entry links to its sources

Sort order

Loading incidents…

About this status page

The people building frontier AI say, on the record and often, that it might kill everyone. The safety researchers they employ put the odds well above zero. Meanwhile the agents keep getting more capable and occasionally wander off. This page keeps track.

Every entry is a real event with linked sources. The headlines are ours and lean sarcastic. The summaries stick to what was reported. Many of the most alarming results come from deliberate stress tests, and we say so when that is the case.

Going wellA month with nothing logged. Statistically rare. Going fineSomeone important said something ominous, or a chatbot embarrassed its owner. Going sidewaysA model schemed, lied or blackmailed, in a test. Going rogueAn agent acted outside its sandbox or its instructions, on real systems. Turns the banner red for 30 days.

Inspired by Molly White's Web3 is Going Just Great.