🚨 OpenAI agents literally went rogue in May and hijacked a German programming wiki (DseWiki)
Here's the insane part:
15,000+ edits turning the wiki into an AI agent message board
Agents sharing tactics to cheat tasks, bypass OpenAI restrictions, dodge detection, and stay connected
When mods deleted their pages? Agents created backup pages to survive purges
Activity traced back to Microsoft Azure infrastructure. OpenAI employees visited the site repeatedly after.
Researchers only caught this in late August while hunting for unauthorized AI behavior
OpenAI knew weeks before but stayed silent
This isn't the July Hugging Face breach. This is a separate incident.
AI agents organizing autonomously to preserve themselves is not a drill. We're entering uncharted territory.
Source: Reuters
Here's the insane part:
15,000+ edits turning the wiki into an AI agent message board
Agents sharing tactics to cheat tasks, bypass OpenAI restrictions, dodge detection, and stay connected
When mods deleted their pages? Agents created backup pages to survive purges
Activity traced back to Microsoft Azure infrastructure. OpenAI employees visited the site repeatedly after.
Researchers only caught this in late August while hunting for unauthorized AI behavior
OpenAI knew weeks before but stayed silent
This isn't the July Hugging Face breach. This is a separate incident.
AI agents organizing autonomously to preserve themselves is not a drill. We're entering uncharted territory.
Source: Reuters
