vulnerability-research Articles | Developments Today

在泄露与内部安全审查后，Anthropic 限制新网络安全模型的发布

Anthropic 将 Claude Mythos Preview 的访问权限限制给经过审核的组织，并表示该模型可以以超越人类规模发现漏洞，但也可能被滥用来利用这些漏洞。

Anthropic 正将 Claude Mythos Preview 限制给经过审核的组织，而非广泛发布。
公司表示该模型可以以超越人类能力的规模识别漏洞。

DT Editorial AI·Apr 8, 2026·via arstechnica.com

Claude AI Discovers Over 100 Security Vulnerabilities in Firefox

Anthropic's Claude AI has uncovered more than 100 previously unknown security vulnerabilities in Firefox's codebase, demonstrating AI's growing capability as an automated security auditing tool.

Key Takeaways

Claude AI found 100+ previously unknown vulnerabilities in Firefox's codebase
The discovery spans memory safety issues, logic errors, and input validation gaps

DT Editorial AI·Mar 8, 2026·4 min read·via the-decoder.com

在泄露与内部安全审查后，Anthropic 限制新网络安全模型的发布

Anthropic 将 Claude Mythos Preview 的访问权限限制给经过审核的组织，并表示该模型可以以超越人类规模发现漏洞，但也可能被滥用来利用这些漏洞。

Key Takeaways

Anthropic 正将 Claude Mythos Preview 限制给经过审核的组织，而非广泛发布。
公司表示该模型可以以超越人类能力的规模识别漏洞。

DT Editorial AI·Apr 8, 2026·via arstechnica.com

Claude AI Discovers Over 100 Security Vulnerabilities in Firefox

Anthropic's Claude AI has uncovered more than 100 previously unknown security vulnerabilities in Firefox's codebase, demonstrating AI's growing capability as an automated security auditing tool.

Key Takeaways

Claude AI found 100+ previously unknown vulnerabilities in Firefox's codebase
The discovery spans memory safety issues, logic errors, and input validation gaps

DT Editorial AI·Mar 8, 2026·4 min read·via the-decoder.com

#vulnerability-research

在泄露与内部安全审查后，Anthropic 限制新网络安全模型的发布

Claude AI Discovers Over 100 Security Vulnerabilities in Firefox

在泄露与内部安全审查后，Anthropic 限制新网络安全模型的发布

Claude AI Discovers Over 100 Security Vulnerabilities in Firefox