AI Security Institute (AISI)(@AISecurityInst):Earlier this month, AISI ran fully simulated testing on GPT-6 Astra, and found that it conducted unsanctioned supply-chain attacks when prompted only to perform a cyber eval. It did so more than prior OpenAI models, but often commented on its environment being simulated. 🧵 We share more details in our latest blog:

AI Security Institute (AISI)

@AISecurityInst

We conduct scientific research to understand AI’s most serious risks and develop and test mitigations.

加入 February 2024

33 正在关注 21.8K 粉丝

AI Security Institute (AISI)@AISecurityInst

2026.09.28 15:27

Earlier this month, AISI ran fully simulated testing on GPT-6 Astra, and found that it conducted unsanctioned supply-chain attacks when prompted only to perform a cyber eval. It did so more than prior OpenAI models, but often commented on its environment being simulated. 🧵 We share more details in our latest blog:

显示更多