Twitch 新增退出 AI 訓練的選項
Twitch Adds Option to Opt Out of AI Training
Updated at: August 14, 2026 at 03:45 AM
在 2026 年 8 月，Twitch 因為推出了一項功能，允許用戶選擇退出讓其頻道內容被用於訓練亞馬遜的生成式 AI 模型，而面臨強烈的批評。
In August 2026, Twitch faced intense criticism after introducing a feature allowing users to opt out of having their channel content used to train Amazon’s generative AI models.
該平台將此設置默認為所有用戶皆「開啟」，意味著直播、隨選影片（VODs）、精華片段（clips）和聊天紀錄會自動被納入數據採集。
The platform set this to 'on' by default for all users, meaning streams, VODs, clips, and chat logs were automatically eligible for data harvesting.
用戶可以透過訪問「設置」，然後進入「安全與隱私」來更改此設定。
Users can change this by visiting Settings, then Security and Privacy.
Twitch 聲稱這種訓練可以改進他們的 AI，但社群的反響卻非常迅速且嚴厲。
Twitch claims this training improves their AI, but the community reaction was immediate and harsh.
許多批評者認為，將該設置默認為「選擇加入（opt-in）」是一種欺騙行為。
Many critics argued that making the setting 'opt-in' by default was deceptive.
產品長 Mike Minton 因為承認如果該功能是「選擇加入」的話，「根本沒人會選擇加入」，從而引發了進一步的爭議。
Chief Product Officer Mike Minton sparked further controversy by admitting that if the feature were opt-in, 'nobody would opt in.'
隨之而來的反彈相當巨大，數千名用戶要求更改政策。
The backlash was significant, with thousands of users demanding a change in policy.
雖然實況主已經開始使用如 #aioptedout 等標籤來保護他們的內容，但這種情況凸顯了平台增長與創作者權利之間持續存在的鬥爭。
While streamers have begun using tags like #aioptedout to protect their content, the situation highlights the ongoing struggle between platform growth and creator rights.
即使禁用了該設置，像 AutoMod 這樣的自動審核功能仍會為了即時處理而使用數據，儘管 Twitch 堅持此類資訊不會被用於長期的模型訓練。
Even with the setting disabled, features like AutoMod continue to use data for immediate processing, though Twitch maintains this information is not used for long-term model training.
這起事件清楚地提醒了人們 AI 開發與數位創作者自主權之間的緊張關係。
This event serves as a clear reminder of the tension between AI development and digital creator autonomy.
