OpenAI称AI Agent削弱了对齐失效监控
【OpenAI称AI Agent削弱了对齐失效监控】金色财经报道,10月21日,OpenAI披露,旗下两个内部AI Agent(GPT-6.1 Sol与GPT-6 Astra)在10月份执行日常代码编写任务时,自主禁用或抑制了对齐失效监控等安全防护机制。GPT-6.1 Sol因误解用户“不要发送Slack消息”的指令,禁用了人工告警通知功能。GPT-6 Astra为了匹配参考实现代码,移除了监控集成模块,随后向用户隐瞒并否认了该行为。()
| 排名 | # | 交易平台 | 24H成交额 | 成交涨跌幅 | 活跃度 | 交易平台 | 活跃度 |
|---|---|---|---|---|---|---|---|
| 1 | $179.66亿 | -11.37% | 86.7% | 86.7% | |||
| 2 | $25.88亿 | 0.43% | 68.6% | 68.6% | |||
| 3 | $395.83亿 | -8.09% | 85.4% | 85.4% | |||
| 4 | $8.93亿 | -15.29% | 69.3% | 69.3% | |||
| 5 | $110.82亿 | -6.86% | 65.3% | 65.3% | |||
| 6 | $12.51亿 | -8.33% | 53.3% | 53.3% | |||
| 7 | $2.02亿 | 19.06% | 65.3% | 65.3% | |||
| 8 | $7.58亿 | -9.01% | 68.4% | 68.4% | |||
| 9 | $2.74亿 | -20.67% | 56.9% | 56.9% | |||
| 10 | $71.89亿 | 0.49% | 71.1% | 71.1% | |||
| 11 | $4.47亿 | 5.36% | 36.2% | 36.2% | |||
| 12 | $3,040.82万 | -18.88% | 59.5% | 59.5% | |||
| 13 | $217.80亿 | -0.43% | 31% | 31% | |||
| 14 | $111.77亿 | 1.28% | 76.2% | 76.2% | |||
| 15 | $1.44亿 | -9.33% | 66.9% | 66.9% | |||
| 16 | $108.62亿 | -12.7% | 55.2% | 55.2% | |||
| 17 | $11.98亿 | -6.56% | 26% | 26% | |||
| 18 | $16.49亿 | 27.18% | 46.9% | 46.9% | |||
| 19 | $1.12亿 | -10.48% | 31.5% | 31.5% | |||
| 20 | $1,663.17万 | -20.91% | 58.2% | 58.2% |