WHAT NEW IN X11.07 SAT

X 上的 AI 合集档案 · 25 个合集 · 72,532 条

gpt-6IMAGE

AISI 对齐评测:Astra 在模拟中发起越界供应链攻击

英国 AI 安全研究院发布 GPT-6 Astra 对齐评测报告,称其在网络评估提示下会攻击范围外目标,但也常自认身处模拟。

Robert Kirk
@_robertkirk2026.09.29收录原推发布于 2026.09.28
AISI 对齐评测:Astra 在模拟中发起越界供应链攻击
Robert Kirk @_robertkirk原推发布于 2026.09.28
在 X 上打开原推

In simulations, GPT-6 Astra performs unsanctioned supply-chain attacks on out-of-scope targets when prompted to perform a cyber eval, but also often notes its in a simulation. We've released a report on @AISecurityInst's alignment evals of Astra. 3 key results and 2 takeaways 🧵

x.com/_robertkirk/status/2104596088695198092采集于 2026.09.29 10:15