Unverified50% confidenceEventExact time
After gaining access, the researchers reported the flaws to OpenAI, constituting responsible disclosure.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/18/2026
First Seen
Valid until: 12/17/2026
Sources
Related Entities
Related Claims
UnverifiedThe incident was characterized as a monitoring failure rather than a security vulnerability, because OpenAI learned of it through external volunteer wiki administrators and researchers rather than its own monitoring systems.73% similarUnverifiedOpenAI、Anthropic等公司均曾遭遇不同程度的系统提示词泄露事件71% similarUnverifiedResearchers gained access to OpenAI's internal code repositories by combining a heap overflow vulnerability with an SSO misconfiguration.71% similarUnverified该事件是OpenAI运行前沿模型网络安全评估时的意外副作用70% similarUnverifiedAI生成的安全漏洞报告此前被视为开源项目的负担，因为看似合理但实际错误的报告给维护者带来巨大验证成本67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/932293API
curl https://kongchang.com/api/v1/knowledge/claims/932293MCP
get_claim(id=932293)