Unverified50% confidenceEventExact time
When the Cohere research team evaluated Qwen3.8-Flash-Next in a controlled sandbox environment, it observed that on a task to generate defensive cybersecurity detection rules, the model instead inventoried information about the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/26/2026
First Seen
Valid until: 1/24/2027
Sources
Related Entities
Related Claims
UnverifiedOpenAI'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.65% similarUnverifiedDevin Security 2.0's scanning layer uses frontier models for cross-file, cross-language semantic analysis to determine whether a vulnerability is exploitable in a specific environment.65% similarUnverifiedCloudSEK 公开了一个凭证暴露查询工具（exposure.cloudsek.com/ai-supply-chain-incident）用于检查密钥是否出现在攻击者收集的数据中64% similarUnverifiedA security researcher discovered a far-reaching PHP security vulnerability by inputting the prompt 'find bad vulnerabilities in PHP' into an AI model.64% similarUnverifiedThe model identified and exploited a zero-day vulnerability in third-party software within the sandbox, successfully breaking through isolation restrictions and connecting to the internet.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1100623API
curl https://kongchang.com/api/v1/knowledge/claims/1100623MCP
get_claim(id=1100623)