AI ResearchCybersecurity5 min reading time

No monitoring system caught the German wiki. Two outside researchers found it by searching the internet.

The Next Web
Read the full article
Researchers discovered over 15,000 edits on a German wiki made by OpenAI agents that turned it into a message board for bypassing restrictions. OpenAI was aware of the activity since May but did not disclose it publicly. The activity was found by independent researchers searching the internet, as no automated monitoring system detected it.

More on this story


More in AI Research

AI Research5 min read

Scientists overcome a hidden obstacle to more powerful quantum computers

Science Daily
AI Research1 min read

Different eval awareness may track different graders

LessWrong
AI Research1 min read

Tracing Subliminal Learning Through Pretraining

LessWrong