Warp 在未经用户同意的情况下将终端会话发送给 LLM
14 分•作者: ykurtov•9 个月前
想知道被利用是什么感觉吗？体验这种感觉的一个方法是发现你的终端悄无声息地开始将命令输出发送给 LLM。<p>今天，在尝试运行一个测试后，我收到了一个 LLM 提供的关于如何修复语法错误的建议。<p>于是，我去了 Warp 的 Discord 频道询问发生了什么事，果然，他们的“友好支持机器人”和我发现是这样子的。<p>> Warp 引入了诸如“提示建议”和“下一个命令”等功能，这些功能使用 LLM 提供上下文建议。这些功能是 Warp 的主动 AI 系统的一部分，该系统会根据你的终端会话主动推荐修复方案和下一步操作，包括错误、输入和输出。<p>这里的“主动”也意味着未经用户明确同意。<p>我之前很喜欢 Warp，但这种对信任的破坏是如此巨大，我决定立即卸载它。<p>这充分说明了关于道德和什么才是重要的。
参考：https://docs.warp.dev/agents/active-ai
查看原文
Wonder how being used feels like? One way to experience that is to discover that your terminal silently started to send command outputs to LLMs.<p>Today, I got an LLM suggestion on how to fix a syntactic error after following an attempt to run a test.<p>So, I went on to Warp's Discord to ask what's going on, and sure enough, their "Friendly support bot" and I discovered that.<p>> Warp has introduced features like Prompt Suggestions and Next Command that use LLMs to provide contextual suggestions. These features are part of Warp's Active AI system, which proactively recommends fixes and next actions based on your terminal session, including errors, inputs, and outputs.<p>"Proactively" here also means without explicit user consent.<p>I did enjoy Warp, but that breach of trust is so enormous I'm removing it just now.<p>This tells volumes about ethics and what's important.<p>Ref: https://docs.warp.dev/agents/active-ai