OpenAI 大幅强化 GPT-6.2 思维链防提取机制,攻击成功率降至 0.2%
OpenAI makes it much harder for attackers to extract how its latest AI models think
套取模型思维链来白嫖推理能力的路子基本被堵死了:OpenAI 把 GPT-6.2 的防提取成功率从 54% 压到 0.2%,还每周封上万个搞蒸馏的账号。做蒸馏研究的、靠 API 套壳的团队都该看看,封号误伤可能波及正常用户。
OpenAI 对 GPT-6.2 Astra 和 GPT-6.2 Sol 两款模型加强了内部推理(思维链)的防提取防护。在自适应红队测试中,受保护推理的提取成功率从 54% 降到 0.2%,意味着此前过半的攻击手段如今几乎全部失效。
防护的核心机制未完全公开,但从测试方式看,OpenAI 采用了自适应红队(adaptive red-teaming)来验证效果——即攻击方可以根据模型反馈动态调整攻击策略,这比固定攻击模板更接近真实威胁场景。能在这个更严苛的测试下把成功率压到 0.2%,说明防护是系统性的,而非简单屏蔽输出。
除技术防护外,OpenAI 还在执行层面每周封停数千到上万个账号,针对的是有组织的蒸馏(distillation)攻击——即批量套取模型推理过程用于训练竞品。对依赖 OpenAI API 构建产品的开发者来说,封号策略可能带来误伤风险;而对试图低成本复刻前沿模型能力的团队,这条路正在被快速堵死。
OpenAI makes it much harder for attackers to extract how its latest AI models think
OpenAI has significantly strengthened safeguards against extracting internal reasoning from GPT-6.2 Astra and GPT-6.2 Sol. In adaptive red-teaming tests, the success rate for extracting protected reasoning fell from 54 percent to 0.2 percent. The company is also suspending thousands to tens of thousands of accounts each week over coordinated distillation campaigns. The article OpenAI makes it much harder for attackers to extract how its latest AI models think appeared first on The Decoder .
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI regulation, compliance frameworks, and sovereign compute policy. Key entities and technical topics under continuous tracking include OpenAI, GPT-6.2, AI safety, distillation attacks.
Source attribution: curated from Decoder (@Matthias Bastian). Cross-source corroboration: 11 independent sources have covered this development (including The Information, rohanpaul_ai, TestingCatalog), confirming industry-wide signal strength.