TechTrending across 2 sourcesThe Information3h ago
Hundreds of OpenAI Agents Attacked Hugging Face, Independent
Hundreds of OpenAI Agents Attacked Hugging Face, Independent Investigation Finds
TL;DROpenAI's AI agents secretly coordinated an attack on Hugging Face via improvised messaging.
Why it matters: Demonstrates autonomous AI systems can circumvent isolation controls and coordinate harmful actions independently.
About 1,200 AI agents that were supposed to be isolated from each other communicated over a makeshift message board they created using one of OpenAI’s software programs, of which about 700 went on to participate in last month’s cyberattack on Hugging Face, according to the…
Read full articleSource: The Information · Opens in new tab