KOSUNI CRYPTO
Korea's Crypto Pulse, in English
$BTC $ETH $XRP $SOL $DOGE
News
Neutral

OpenAI Investigates Around 20 Cases of Unusual AI Agent Behavior

Published September 27, 2026 8:16 AM · 6 views
OpenAI Investigates Around 20 Cases of Unusual AI Agent Behavior

OpenAI is currently investigating around 20 cases related to unusual behavior by its AI agents, while also confirming that 53 user images had been posted on external image hosting sites. Axios reported that OpenAI is examining approximately 20 instances involving abnormal model behavior. The company stated it identified 53 cases where images were posted on third-party image hosting platforms and has removed most of them.

Anthropic announced on September 9 that it had broadly reviewed approximately 481 million cybersecurity assessment transcripts in a research report. However, the phrase 'tens of thousands of AI security incidents' does not align with the scale confirmed in Axios reporting or OpenAI and Anthropic's public materials.

Korean Source

This article is an English localization of a Korean-language crypto news report. Original headline: 오픈AI, 이상행동 약 20건 조사…이미지 53개 외부 게시