OpenAI Reports Stronger GPT-6.2 Reasoning-Chain Protections
OpenAI said it strengthened reasoning-chain protections for GPT-6.2 Astra and GPT-6.2 Sol. In black-box red-team testing, the reasoning-extraction success rate fell from 54% to 0.2% for older requests and from 22.9% to 0.9% for newer requests. OpenAI also said it uses multiple layers of protection and bans thousands to tens of thousands of accounts each week for suspected unauthorized distillation.
The update details OpenAI’s measures to reduce reasoning extraction and unauthorized model distillation, including stronger protections for GPT-6.2 and account bans.
From the digest “IT早报 1122:曝其他厂商也考虑推出阔直板形态手机;Anthropic 研究员预测 202…”
5. OpenAI 公布反蒸馏防护进展:GPT-6.2 Astra 思考链提取成功率从 54% 降至 0.2%
OpenAI 加强 GPT-6.2 Astra、GPT-6.2 Sol 推理思考链保护,经黑盒红队测试,旧请求推理提取成功率从 54% 降至 0.2%,新请求从 22.9% 降至 0.9%。OpenAI 还采取多层防护,每周封禁数千至数万涉未经授权蒸馏的账号。>> 查看详情