Agent权限记忆篡改导致授权洗白，修复需验证写入
An agent can make a perfectly consistent authorization decision and still be wro…
Agent安全领域的关键漏洞揭示，给出了具体的故障率数据和可落地的修复方案，做Agent架构的同学必读。
An agent can make a perfectly consistent authorization decision and still be wrong if its memory has already rewritten who is allowed to do what.
如果代理的记忆已经重写了谁被允许做什么，那么即使它做出了完全一致的授权决策，也可能是错误的。
This paper studies that failure in long-running agents, where permissions, revocations, and scope changes get compressed into persistent memory.
本文研究了长期运行代理中的这种故障，其中权限、撤销和范围变更被压缩到持久化记忆中。
A revoked or narrowed permission can be remembered incorrectly, and that bad memory becomes the executor’s version of the truth.
被撤销或缩小的权限可能会被错误地记住，而这种糟糕的记忆会成为执行者的事实版本。
In the paper’s typed incremental-memory tests, false authority appeared for up to 50.2% of unauthorized requests. Once that false permission existed, executors acted on it in 98.6% of matched trials.
在论文的带类型增量记忆测试中，高达 50.2% 的未授权请求出现了虚假授权。一旦存在该虚假权限，执行者在 98.6% 的匹配试验中都会依据其采取行动。
Changing only the memory to the exact authorization state dropped those unauthorized actions to 0%. That points to memory, not the executor, as the main failure point in these cases.
仅将记忆更改为精确的授权状态，可将这些未授权操作降至 0%。这表明在这些案例中，主要故障点在于记忆而非执行者。
The fix is to verify permissions before writing them into memory, preserve where each permission came from, and track grants, changes, and revocations with deterministic state updates.
修复方法是：在将权限写入记忆之前验证权限，保留每个权限的来源，并通过确定性状态更新来跟踪授予、变更和撤销。
– arxiv. org/abs/2609.01836
– arxiv.org/abs/2609.01836
Title: "Agent Memory Is a Surface for Endogenous Authorization Laundering"
标题：“代理记忆是内生授权洗白的表面”
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力