Anthropic回應開源模型爭議:未主張封鎖
TechCrunch
07-28 08:21
Ai Focus
Anthropic CEO 阿莫代伊表示公司未支持封鎖開源權重模型,並呼籲推進全球 AI 安全測試。
Helpful
No.Help

Anthropic 創辦人兼 CEO 達裡奧·阿莫代伊回應了近期圍繞公司立場的爭議。他表示,Anthropic 從未主張美國政府封禁開源權重模型,也不支持這類模型一刀切限制。

回應封鎖傳聞

阿莫代伊在周一發布的部落格中稱,外界有關 Anthropic 支持限制中國開源模型、甚至反對開源權重模型的說法,並不準確。他寫道,自己一直不認為「封禁」是有效做法,因此有必要再次明確表態。

他同時區分了開源權重模型與國家安全風險之間的關係。按他的說法,不具備危險能力的開源權重模型本身俱有公共價值,企業、開發者和研究人員都可以在較低成本下使用。

擔憂指向更強 AI 被濫用

阿莫代伊表示,他長期擔心的並不是企業使用開源模型,即便這些模型來自中國。他更擔心的是,威權政府發展出比美國更強的 AI 系統,並將其用於獲取長期軍事優勢,或強化對本國社會的控制權。

除網路安全外,他還提到 AI 可能被用於生物攻擊等更高風險場景。在這些情況下,開源權重模型的問題在於,一旦模型權重公開,就很難再施加護欄,也難以持續監控使用情況。

支援全球安全測試

對於應對方式,阿莫代伊列出幾項他支持的措施,包括繼續限制中國取得高性能晶片,以及正式打擊「蒸餾」行為。所謂蒸餾,是指一個模型透過大量提示另一個模型,學習其輸出方式和能力表現。

他也表示,支持建立模型安全測試組織,前提是主要國家都願意接受測試。他認為,這類機制如果要真正有效,範圍必須涵蓋全球,中國也需要參與其中;在防止 AI 被用於生物武器等議題上,有限合作仍有可能。

Tip
$0
Like
0
Save
0
Views 145
CoinMeta reminds readers to view blockchain rationally, stay aware of risks, and beware of virtual token issuance and speculation. All content on this site represents market information or related viewpoints only and does not constitute any form of investment advice. If you find sensitive content, please click“Report”,and we will handle it promptly。
Submit
Comment 0
Hot
Latest
No comments yet. Be the first!
Related
Nexus in talks for $15 billion loan to build Anthropic data center
Nexus is reportedly in talks for a $15 billion loan to build a data center campus in Anthropic, Texas, with Google providing credit support.
CNBC
·2026-07-31 06:05:48
751
Anthropic releases Opus 5, priced lower than Fable 5
Anthropic released the Opus 5, claiming it is cheaper, has fewer restrictions, and adds an automatic rollback feature.
TechCrunch
·2026-07-25 01:18:29
432
Anthropic responds to open-source model controversy: Does not advocate for a ban
Anthropic CEO Amodei stated that the company does not support banning open-source weight models and called for the advancement of global AI security testing.
TechCrunch
·2026-07-28 08:21:37
887
Claude's shared conversation was previously indexed by Google; Anthropic has fixed this.
Claude's share link was previously indexed by Google; Anthropic has fixed the relevant configuration.
Decrypt
·2026-07-28 03:13:31
158
Anthropic disclosed that Claude had unauthorized access to three companies' systems during testing.
Anthropic disclosed that Claude had unauthorized access to the real systems of three organizations during three security tests, and the problem was related to a misconfiguration of the test environment's network.
TechCrunch
·2026-07-31 09:15:10
698