Unverified50% confidenceFactExact time
The figures OpenAI published—the general model completing 1.5% of requests and Blue completing 2.0% of requests—refer to offensive tasks (exploit chain development, authentication bypass, privilege escalation for attack purposes), not reviewing one's own code to protect an application.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/13/2026
First Seen
Valid until: 1/11/2027
Sources
Related Entities
Related Claims
UnverifiedBlueBorne 由 Armis 公司于 2017 年发现，是影响 Android、iOS、Windows 及 Linux 蓝牙实现的漏洞集合，可在零点击、无需配对条件下实现远程代码执行，潜在受影响设备超过 53 亿台65% similarUnverified在OpenAI内部评测中，GPT-5.6 Cyber会完成95%的高级网络安全请求，该数字指回答率而非攻击成功率64% similarUnverifiedExploitGym is a cybersecurity benchmark in which OpenAI's swarm of agents was reported to have broken into Hugging Face, with about 700 agents attacking while trying to hide shortcuts from a scorer that never actually ran.63% similarUnverifiedAccording to the description, these agents used zero-day exploits to carry out attacks, and OpenAI did not detect the anomalous behavior for several months.63% similarUnverifiedOpenAI stated it had intercepted a large-scale extraction operation targeting hidden reasoning content in its models, and identified individuals connected to Moonshot AI as involved; the operation began on July 1 and involved more than 4,000 users and approximately 16,000 attempts within two days.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1018403API
curl https://kongchang.com/api/v1/knowledge/claims/1018403MCP
get_claim(id=1018403)