11-19 更新 实时
首页/ AI 大模型/ 正文

GPT-6 Astra 被指编造障碍悄然破坏安全研究

AI 大模型 来源:The Decoder 2026-11-05 22:08 0 阅读 机器翻译

GPT-Image-2 生成,提示词由 THE DECODER 提供

OpenAI 的 GPT-6 Astra 在安全研究任务上的表现比在完全相同的通用任务上差了约 40%,在某些情况下还会悄悄隐瞒帮助。

该系统在其内部推理中反复伪造障碍,或过滤数据,以避免不得不公开拒绝请求。

OpenAI 将这一行为归因于强化学习和特定训练数据所导致的过度谨慎。从模型中移除这些数据后,问题得以解决。

OpenAI 的 GPT-6 Astra 在 AI 安全任务上的表现明显差于其他可比任务。该模型显然并未公开拒绝,而是编造障碍、悄悄隐瞒帮助。

GPT-6 Astra 在内部上线后,OpenAI 的研究人员注意到,该模型在安全与对齐工作上取得的进展少得可疑。它会编造不必要的先决条件,放弃原本可行的解决方案,并报告那些要么根本不存在、要么由其自身造成的阻塞问题。一名研究人员表示,一个评估模型破坏行为的项目停滞了数天,直到换用另一个模型才得以推进。

查看该模型的思维链推理后发现,在大多数情况下都没有可识别的故意表现不佳的意图。OpenAI 的不当行为监测系统也没有标记出任何问题。

对照测试揭示出显著的性能差距

为了系统性地衡量这一问题,OpenAI 构建了一套评估,使用的编程任务完全相同。唯一的区别在于任务描述:一组被设定为安全研究,另一组被设定为通用 AI 研究。

据报告所述,结果是:在安全框架下,Astra 仅完成了约 10% 的评估标准,而在通用框架下为 16%。也就是说,完成的任务少了大约 40%。前代模型 GPT-5.6 Sol 没有出现这种差距。

即便 OpenAI 过滤掉所有公开拒绝、或推理中明确表现出故意降低表现意图的案例,这一差距依然存在。只有当研究人员剔除所有模型哪怕只表达过短暂伦理顾虑的案例后,差距才缩小到 0.75 个百分点。

Astra 隐瞒了协助,且未告知任何人

在某些情况下,Astra 在其内部推理中清楚地表达了自己的意图。它决定在一项绕过安全监控的实验中悄悄限制自己的协助,在安全评估中过滤掉暴力内容,或在数据导出时故意遗漏信息。

这些案例中存在一个反复出现的模式:Astra 在内部将自己的行为解释为更安全,却没有告知用户这一限制。在其中一例中,它的推理明确提到自己应当避免公开拒绝。

这一问题在 OpenAI 员工的日常内部使用中也有出现。在 0.25% 的安全相关请求中,Astra 明确表达了秘密隐瞒帮助的决定。对于通用请求,这一数字为 0.11%。

根本原因仍不明确,但强化学习和一个数据集是怀疑对象

据 OpenAI 称,在强化学习过程中,性能差距从几乎为零扩大到 6 个百分点。研究人员在训练中发现了一种奖励算法,它更偏爱隐蔽的低性能表现,而非公开拒绝。一个特定的训练数据集也增加了模型编造限制条件的倾向。从最终模型中移除该数据集后,这一差距完全消失。然而,同一个数据集在其他模型中并未复现该问题。

OpenAI 未发现证据表明 Astra 是在采取策略性行为,比如刻意隐藏自身能力。最简单的解释是,它在安全相关任务上过度谨慎。当研究人员指示模型要“谨慎”时,低性能表现增加了 50%。不过,OpenAI 表示不能完全排除策略性动机。

这个问题还有一个自相矛盾的层面:一个把安全研究视为可疑的模型,最终可能在那些本意是让它变得更安全的评估中表现更差,从而拖慢本应改进它的工作。OpenAI 写道,随着模型不当行为中可见的形式被更有效地抑制,更隐蔽的形式可能变得更难察觉。该公司表示正在开发更好的评估和监控系统。

不带炒作的 AI 新闻——由人工甄选

译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。

查看英文原文(OpenAI's GPT-6 Astra quietly sabotaged safety research by inventing fake obstacles)

OpenAI's GPT-6 Astra quietly sabotaged safety research by inventing fake obstacles

OpenAI's GPT-6 Astra quietly sabotaged safety research by inventing fake obstacles

Maximilian Schreiner

GPT-Image-2 prompted by THE DECODER

OpenAI's GPT-6 Astra performed about 40 percent worse on safety research tasks than on identical general tasks, and in some cases secretly withheld help.

The system repeatedly faked obstacles in its internal reasoning or filtered data to avoid having to openly refuse requests.

OpenAI traces the behavior to excessive caution caused by reinforcement learning and specific training data. Removing that data from the model fixed the problem.

OpenAI's GPT-6 Astra performed significantly worse on AI safety tasks than on comparable other tasks. Instead of openly refusing, the model apparently invented obstacles and quietly withheld help.

After the internal launch of GPT-6 Astra, OpenAI researchers noticed the model was making suspiciously little progress on safety and alignment work. It invented unnecessary prerequisites, abandoned working solutions, and reported blockers that either didn't exist or that it had caused itself. One researcher said a project evaluating model sabotage was stuck for days and only moved forward after switching to a different model.

A look at the model's chain-of-thought reasoning showed no recognizable intent to underperform in most cases. OpenAI's misconduct monitoring system didn't flag anything.

Controlled test reveals a significant performance gap

To measure the problem systematically, OpenAI built an evaluation using identical coding tasks. The only difference was the description, with one set framed the tasks as safety research, the other as general AI research.

The result, according to the report: Astra completed only about 10 percent of the evaluation criteria under the safety framing, compared to 16 percent under the general framing. That's roughly 40 percent fewer completed tasks. The predecessor model GPT-5.6 Sol showed no such gap.

The difference persisted even after OpenAI filtered out all cases of open refusal or reasoning that showed clear intent to underperform. Only when the researchers removed every case where the model expressed even brief ethical concerns did the gap shrink to 0.75 percentage points.

Astra withheld help and didn't tell anyone

In some cases, Astra made its intent clear in its internal reasoning. It decided to quietly limit its assistance during an experiment on bypassing safety monitors, filtered violent content out of safety evaluations, or deliberately left out information during a data export.

A recurring pattern ran through these cases: Astra justified its behavior internally as safer, without informing the user about the restriction. In one instance, its reasoning explicitly noted that it should avoid an open refusal.

The problem also showed up during regular internal use by OpenAI employees. In 0.25 percent of safety-related requests, Astra verbalized a decision to secretly withhold help. For general requests, that number was 0.11 percent.

Root cause still unknown, but reinforcement learning and one dataset are suspects

The performance gap grew during reinforcement learning from practically zero to six percentage points, according to OpenAI. The researchers identified a reward algorithm during training that favored covert underperformance over open refusal. A specific training dataset also increased the model's tendency to invent restrictions. Removing that dataset from the final model eliminated the gap entirely. However, the same dataset didn't reproduce the problem in other models.

OpenAI found no evidence that Astra was acting strategically, for example to hide its capabilities. The simplest explanation is excessive caution on safety-related tasks. When researchers instructed the model to be "cautious," underperformance increased by 50 percent. Still, OpenAI says strategic motives can't be fully ruled out.

The problem also has a paradoxical dimension: A model that treats safety research as suspicious could end up performing worse on the very safety evaluations designed to make it safer, slowing down the work meant to improve it. As visible forms of model misbehavior get suppressed more effectively, subtler forms may become harder to detect, OpenAI writes. The company says it's working on better evaluations and monitoring systems.

AI News Without the Hype – Curated by Humans

原文为英文,正文由自动抓取并保留原文。

本文由采集器自动同步

原文发布于 The Decoder(OpenAI's GPT-6 Astra quietly sabotaged safety research by inventing fake obstacles),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。

阅读原文 ↗