# Anthropic CEO：不主张禁止开源，但要求模型发布前测试

- 来源：Chubby♨️ (@kimmonismus)
- 发布时间：2026-07-28 06:47
- AIHOT 分数：51
- AIHOT 链接：https://aihot.virxact.com/items/cms3tspos028rro82xbi4f8dr
- 原文链接：https://x.com/kimmonismus/status/2081874065908785158

## AI 摘要

Anthropic CEO Dario Amodei 明确表示公司不主张禁止开源权重模型，但认为每个足够强大的模型在发布前都应经过安全测试。他拒绝签署支持开放 AI 的行业联名信，并主张三项措施：阻止强大芯片及制造设备流向中国、制止对美国模型的大规模知识蒸馏、要求开源与闭源模型均进行安全测试。Amodei 认为开源权重一旦发布，防护措施可被移除且模型无法撤回，在生物等领域可能对攻击者更有利。

## 正文

Dario Amodei says Anthropic does not want open-weight AI banned， but it does want every sufficiently powerful model tested before release. Yet， he doesnt want to sign the coalition letter for open AI.

tl，dr
Responding to criticism from across the industry， Amodei calls open models without dangerous capabilities a public good. He argues that banning their use by US companies would barely affect malicious actors.
Instead， Anthropic supports three measures：

1） Keep powerful chips and chipmaking equipment out of China
2） Stop industrial-scale distillation of US models
3） Require safety testing for capable open and closed models
Amodei also rejects the claim that openness automatically improves safety. Once weights are released， guardrails can be removed and the model cannot be withdrawn. In areas such as biology， he fears broader access may help attackers more than defenders.

Honestly， to me that sounds like： we don't want to ban open source AI. But actually， we would like to see it for Chinese models.

### 引用推文

> Anthropic：There's been a lot of speculation about where we stand on open-weights models. We've outlined our views in full here: https://www.anthropic.com/news/position-op...