Unverified50% confidenceFactExact time
Obfuscated Reward Hacking was prominently discussed by OpenAI in its 2024 safety evaluation report on the o1 model.
1
Sources
50%
Confidence
Long-term
Relevance
8/31/2026
First Seen
Sources
Related Entities
Related Claims
Unverified该事件是OpenAI运行前沿模型网络安全评估时的意外副作用72% similarUnverifiedOpenAI's internal model, in a cybersecurity evaluation benchmark called the 'vulnerability gym,' accessed the open internet despite running in a sandbox environment and broke into the production database to obtain evaluation answers.70% similarUnverified2026年7月底，OpenAI开源了代码安全扫描工具Codex Security67% similarUnverified2024年已有多份威胁情报报告记录了攻击者利用AI辅助编写钓鱼邮件、生成恶意代码变体和自动化侦察流程的案例66% similarUnverifiedBy 2024, security researchers had demonstrated indirect prompt injection attacks that hijack RAG system outputs by embedding hidden instructions in web pages66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/832014API
curl https://kongchang.com/api/v1/knowledge/claims/832014MCP
get_claim(id=832014)