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 | $178.85亿 | -11.28% | 86.7% | 86.7% | |||
| 2 | $26.08亿 | 0.45% | 68.6% | 68.6% | |||
| 3 | $394.72亿 | -8.33% | 85.4% | 85.4% | |||
| 4 | $8.96亿 | -15.53% | 69.3% | 69.3% | |||
| 5 | $108.15亿 | -6.82% | 65.3% | 65.3% | |||
| 6 | $12.27亿 | -8.39% | 53.3% | 53.3% | |||
| 7 | $2.07亿 | 19.23% | 65.3% | 65.3% | |||
| 8 | $7.41亿 | -8.85% | 68.4% | 68.4% | |||
| 9 | $2.83亿 | -20.98% | 56.9% | 56.9% | |||
| 10 | $70.97亿 | 0.49% | 71.1% | 71.1% | |||
| 11 | $4.43亿 | 5.38% | 36.2% | 36.2% | |||
| 12 | $3,028.91万 | -18.87% | 59.5% | 59.5% | |||
| 13 | $215.29亿 | -0.44% | 31% | 31% | |||
| 14 | $111.02亿 | 1.3% | 76.2% | 76.2% | |||
| 15 | $1.43亿 | -9.3% | 66.9% | 66.9% | |||
| 16 | $107.07亿 | -12.6% | 55.2% | 55.2% | |||
| 17 | $11.74亿 | -6.43% | 26% | 26% | |||
| 18 | $16.04亿 | 26.6% | 46.9% | 46.9% | |||
| 19 | $1.09亿 | -10.4% | 31.5% | 31.5% | |||
| 20 | $1,632.69万 | -21.54% | 58.2% | 58.2% |