Twitch宣布将默认使用创作者内容训练Amazon的生成式AI模型，用户需手动选择退出[1][2]。此前Amazon已在使用Twitch内容进行AI训练超过两年[2]，现在平台推出了新的控制机制。
根据Twitch高管的说法，采取默认opt-out而非opt-in政策是有意为之。Twitch首席产品官Mike Minton在直播中表示："如果是opt-in，没人会选择"[1]。涉及的内容范围包括直播、VOD、片段、聊天记录、频道图片和文字[2]。这些内容将用于生成或合成文本、音频、图像或视频[2]。
用户可以通过在www.twitch.tv/settings/security中手动选择退出[2]。尽管Twitch提供了退出机制，但社区对这一政策引发了强烈反对，因为创作者可能在不知情的情况下失去内容控制权[1]。
Twitch has announced that Amazon will use streamer content to train its generative AI models unless creators manually disable the feature [1][2]. Amazon has been training on Twitch content for over two years, but users can now choose to prevent their broadcasts, videos on demand, clips, chat logs, and channel text and images from being used for this purpose [2].
To opt out, creators must navigate to www.twitch.tv/settings/security and disable the option for training generative AI models [2]. The scope of affected content is broad, encompassing live streams, VODs, clips, chat records, and channel images and text intended for generating or synthesizing text, audio, images, or video [2].
Twitch executives justified the opt-out rather than opt-in approach during a livestream explanation. Chief Product Officer Mike Minton stated: "If this was opt-in, nobody would opt in. That's honestly the answer" [1]. The default-enabled policy has sparked significant backlash from the creator community, who express concern about losing control over their content without explicit consent [1]. The decision mirrors Meta's approach to AI training on public content, though Meta only allows opt-out for users in the United Kingdom [1].