Anthropic 向更多安全团队开放低安全限制的 Claude:前身计划只在 2026 年 4 到 7 月就发现 12.9 万个已确认漏洞,其中超过 3.3 万个被评为高危或严重
Anthropic gives more security teams access to Claude with fewer safety restrictions
精选评分 78

导读
Anthropic 正在扩大它的 Cyber Verification Program,让更多安全专业人员以更低的安全限制使用 Claude 模型,用途涵盖渗透测试、恶意软件分析与漏洞研究。公司给出的收益数据是这项计划最有力的部分:在其前身计划中,合作方从 2026 年 4 月到 7 月一共发现至少 12.9 万个已确认的漏洞,其中超过 3.3 万个被评为高危或严重。这条记录的用处在于它把「对可信第三方放开模型限制」这一颇具争议的决策,配上了一段可核查的成果周期——它并不能证明放开总是对的,但确实给了防御性用途一个具体的量化依据。
推荐理由
把「放开安全限制」这件事配上可量化的收益数据(12.9 万确认漏洞、3.3 万高危),是讨论模型能力该向谁放开时少见的实证。
主体与标签
Anthropic#安全/对齐#政策/监管
正文

Anthropic is expanding its Cyber Verification Program, giving more security professionals access to Claude models with fewer safety restrictions for penetration testing, malware analysis, and vulnerability research. Anthropic says partners in its predecessor program found at least 129,000 confirmed vulnerabilities from April through July 2026, including more than 33,000 rated high-severity or critical.
The article Anthropic gives more security teams access to Claude with fewer safety restrictions appeared first on The Decoder .