- OpenAI會在2026年9月28日或之前宣布已恢復訓練嗎？3¢
- OpenAI會在2026年9月29日宣布已恢復訓練嗎？<1¢
- OpenAI會在2026年9月30日宣布已恢復訓練嗎？<1¢
- OpenAI 會在 2026 年 10 月 1 日宣布已恢復訓練嗎？10¢
- OpenAI會在2026年10月2日宣布已恢復訓練嗎？7¢
- OpenAI會在2026年10月3日宣布已恢復訓練嗎？2¢
- OpenAI會在2026年10月4日宣布已恢復訓練嗎？1¢
- OpenAI 會在 2026 年 10 月 5 日宣布已恢復訓練嗎？4¢
- OpenAI會在2026年10月6日宣布已恢復訓練嗎？4¢
- OpenAI會在2026年10月7日宣布恢復訓練嗎？2¢
- OpenAI 會在 2026 年 10 月 8 日宣布已恢復訓練嗎？<1¢
- OpenAI 會在 2026 年 10 月 9 日宣布已恢復訓練嗎？4¢
On September 25, 2026, amid reports that its AI agents had probed U.S. government websites, including those of the Census Bureau and Education Department, OpenAI disclosed a pause in training, evaluation, and inference involving tool use for its most capable models. This market will resolve according to the calendar date (ET) on which OpenAI announces that it has resumed training covered by the pause announced on September 25, 2026. If no qualifying announcement is made by 11:59 PM ET on October 20, 2026, this market will resolve to "Not resumed by October 20". Resuming training means resuming, in part or in full, training suspended under the pause announced on September 25, 2026, including beginning a fresh training run in place of a suspended run. A qualifying announcement must be made through OpenAI's official channels or by an authorized representative acting in an official capacity. The totality of the announcement and its surrounding context must clearly establish that training covered by the pause announced on September 25 has resumed, even if the announcement does not name the incident or specify the date the pause was announced. Statements that training is permitted, planned, or expected to restart do not by themselves qualify. Training that was not subject to the pause announced on September 25, 2026, does not qualify. This market will resolve to the date of the first qualifying announcement, regardless of whether the announcement is later modified or withdrawn, or training is subsequently paused again. The primary resolution source for this market will be official information from OpenAI; however, a consensus of credible reporting may also be used.