Event ·Multi-source

Marcus质疑OpenAI安全报告可靠性

GaryMarcus支持TylerTracy观点，认为OpenAI的安全报告无法证明能有效防止Astra模型规避安全系统。OpenAI使用难以从外部评估的内部评估方法，未详细说明人工升级/响应流程，且未充分测试Astra的攻击和规避能力。OpenAI过去曾低估此类风险，Marcus对其当前安全措施缺乏信心。

Current read

Marcus和Tracy共同质疑OpenAI安全报告的充分性，指出其评估方法不透明且可能低估风险。

11 sources58° AI heatLast updated 2026/9/4 04:39:55

Evidence

11 sources

Primary sourceGary Marcus

2026/9/4 04:39:55View source

Related source 1IT之家

2026/9/3 09:56:59View source

Related source 2Decoder

2026/9/2 14:20:28View source

Related source 3VnExpress Số hóa

2026/9/2 14:58:03View source

Related source 4techcrunch

2026/9/2 20:19:14View source

Related source 5OpenAI Blog

2026/9/3 00:00:00View source

Related source 6The Rundown AI

2026/9/3 10:30:16View source

Related source 7Justine Moore

2026/9/3 20:01:07View source

Related source 8Rappler: Technology

2026/9/4 02:50:08View source

Related source 9Simon Willison

2026/9/4 17:40:35View source

Related source 10Genk

2026/9/5 09:05:00View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story