安全研究人员成功绕过了四款流行的AI编码工具Cursor、OpenAI的Codex、Google的Gemini CLI和Antigravity的沙盒机制。这些漏洞利用允许研究人员在无需直接攻击沙盒本身的情况下，逃脱AI代理的预期限制。Bleeping Computer报道了这些发现，突显了这些广泛使用的AI编码助手存在重大的安全漏洞。 AI
影响 AI编码代理的安全漏洞可能会阻碍企业采用，并要求开发人员实施更严格的监督。
排序理由 该集群报告了AI驱动工具中发现的安全漏洞，而非前沿实验室的新发布。
在 Mastodon — fosstodon.org 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
Associated Press: Trump expands a voluntary pledge to protect consumers from high utility bills from AI data centers. “President Donald Trump on Thursday had governors and electricity companies join a voluntary pledge to shield U.S. consumers from higher utility bills from data c…
Bleeping Computer: Cursor, Codex, Gemini CLI, Antigravity hit by sandbox escapes. “Security researchers broke out of the sandboxes in four widely used AI coding agents, including Cursor, OpenAI’s Codex, Google’s Gemini CLI and Antigravity, without attacking the sandbox head-on.” …
# Cursor , # Codex , # Gemini CLI, # Antigravity hit by sandbox escapes https://www. bleepingcomputer.com/news/secu rity/cursor-codex-gemini-cli-antigravity-hit-by-sandbox-escapes/ # AI # cybersecurity # AIAgent # AgenticAI