Unverified50% confidenceBenchmarkExact time
In cybersecurity vulnerability reproduction and remediation tests, Large 4 scores 82%, while several open-source large models score near zero (due to outright refusal).
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/30/2026
First Seen
Valid until: 11/13/2026
Sources
Related Entities
Related Claims
Unverified根据2026年3月OpenAI产品研究预览公告，过去30天在测试用户群扫描超过120万次提交，发现792个严重漏洞和10561个高危漏洞，官方称假阳性率下降超50%、严重性过度报告降低超90%70% 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68% similarVerifiedThe four essential elements of a vulnerability report are the vulnerability URL, reproduction steps, impact/harm, and remediation suggestions.66% similarUnverifiedThe OpenAI system card shows that in prompt injection tests against GPT-6 Astra, none of 16,600 attacks across 500 simulated emails succeeded; an adaptive attacker's 2,638 attempts had zero success.66% similarUnverifiedThe vulnerability fixed in version 0.65.5 is: when the requested table name contains a trailing newline character at the end, it is possible to bypass table-level permission checks and expose private row data that should be protected.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114958API
curl https://kongchang.com/api/v1/knowledge/claims/1114958MCP
get_claim(id=1114958)