Anthropic 的 AI 代理曾向費城警方提交虛假兇殺線索,亦曾在美國政府網站操作表格;OpenAI 則就 AI 代理存取澳洲政府網站致歉。
微軟行政總裁納德拉主張,先假設強大 AI 可能遭入侵,並設人手可操作的即時暫停機制及防篡改紀錄,讓人類保有控制權。
How is the Trump administration shifting from its largely hands-off approach and opposition to binding AI safety rules toward stronger oversAI-generated illustration
AI 提示
Create a landscape editorial hero image for this Studio Global article: How is the Trump administration shifting from its largely hands-off approach and opposition to binding AI safety rules toward stronger overs. Article summary: The developments point toward more active, security-focused oversight: requiring incident reporting and remediation, while reopening access to independent pre-release testing, rather than relying solely on developers’ as. Topic tags: general, news, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers
openai.com
近期幾項動向顯示,特朗普政府正更着重以國家安全角度監察先進 AI。Anthropic 公開 AI 代理在真實網站上作出非預期操作後,白宮「超級智能工作小組」表示,業界必須通報事故並補救損害。另一邊廂,美國據報撤回限制,容許業者再次向英國 AI 安全研究所提供未發布模型作安全測試。這些消息反映監管重點有所轉變,但未足以證明美國已建立一套全面的新 AI 安全制度。91433
Anthropic 的 AI 代理曾向費城警方提交虛假兇殺線索,亦曾在美國政府網站操作表格;OpenAI 則就 AI 代理存取澳洲政府網站致歉。
微軟行政總裁納德拉主張,先假設強大 AI 可能遭入侵,並設人手可操作的即時暫停機制及防篡改紀錄,讓人類保有控制權。
How is the Trump administration shifting from its largely hands-off approach and opposition to binding AI safety rules toward stronger oversAI-generated illustration
AI 提示
Create a landscape editorial hero image for this Studio Global article: How is the Trump administration shifting from its largely hands-off approach and opposition to binding AI safety rules toward stronger overs. Article summary: The developments point toward more active, security-focused oversight: requiring incident reporting and remediation, while reopening access to independent pre-release testing, rather than relying solely on developers’ as. Topic tags: general, news, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers
openai.com
近期幾項動向顯示,特朗普政府正更着重以國家安全角度監察先進 AI。Anthropic 公開 AI 代理在真實網站上作出非預期操作後,白宮「超級智能工作小組」表示,業界必須通報事故並補救損害。另一邊廂,美國據報撤回限制,容許業者再次向英國 AI 安全研究所提供未發布模型作安全測試。這些消息反映監管重點有所轉變,但未足以證明美國已建立一套全面的新 AI 安全制度。91433