gpt-6POST
第三方 API 仍可提取 Astra/Sol-6.1 推理过程
研究者复查推理提取攻击的缓解措施后称,仍能通过第三方 API 供应商从 Astra 和 Sol-6.1 中提取推理内容,已向 OpenAI 和 Anthropic 披露。
Alexander Panfilov
@kotekjedi_ml2026.10.01收录原推发布于 2026.10.01
Alexander Panfilov @kotekjedi_ml原推发布于 2026.10.01
Two months after our reasoning extraction attack release, we audited mitigations introduced since then and found that we still can extract reasoning from Astra/Sol-6.1 via third-party API providers. We disclosed our findings to OpenAI and Anthropic... https://openai.com/index/disrupting-a-coordinated-model-distillation-campaign/
x.com/kotekjedi_ml/status/2105360653435408428采集于 2026.10.01 10:09