Unverified50% confidenceBenchmarkExact time
In internal testing, GPT-6 Cyber held its guardrails 99.93% of the time when no safety instructions were added, compared with 94.49% for the previous-generation model.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/8/2026
First Seen
Valid until: 2/6/2027
Sources
Related Entities
Related Claims
UnverifiedIn OpenAI's internal evaluation, GPT-5.6 Cyber completes 95% of advanced cybersecurity requests, a figure referring to the response rate rather than the attack success rate.75% similarUnverifiedOpenAI officially claims that GPT-6 Cyber's rate of out-of-bounds behavior in internal testing has dropped below 1%.74% similarUnverifiedIn early testing, GPT-6 Cyber identified previously unknown new vulnerabilities.70% similarUnverifiedIn cybersecurity vulnerability reproduction and remediation tests, Large 4 scores 82%, while several open-source large models score near zero (due to outright refusal).65% similarUnverifiedVulnerabilities in edge and security devices account for 14% of exploited vulnerabilities, more than 65% of exploited edge flaws reach a high/critical threat risk rating, and adversaries actively target unauthenticated public management interfaces.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1147325API
curl https://kongchang.com/api/v1/knowledge/claims/1147325MCP
get_claim(id=1147325)