Google must let publishers opt out of AI Search features, rules UK 英国监管机构要求谷歌允许出版商选择退出AI搜索功能

The UK’s Competition and Markets Authority just handed publishers a slingshot to use against Google’s AI juggernaut. Effective immediately, website owners can block their content from being ingested into AI Overviews or used to fine-tune Google’s models. It’s a world-first regulation, and the CMA is framing it as a victory for negotiation leverage. But let’s be clear: this isn’t a revolution. It’s a tactical pause in a war that’s just getting started.

On the surface, it’s a meaningful win for the publishers who have watched helplessly as their journalism—often the very material that makes Google’s search results valuable—was scraped, summarized, and served back to users in a neat, ad-sorbing box. Giving them the opt-out switch is a matter of basic digital property rights. It forces Google to acknowledge that the content pipeline isn’t a one-way street for the taking. News organizations, in particular, have been screaming about this for years as their business models crumbled under the weight of aggregated, platform-mediated audiences.

But here’s the critical judgment: this regulation treats the symptom, not the disease. The disease is the overwhelming data asymmetry between a platform like Google and the rest of the web. By letting publishers opt out of AI training and overview inclusion, the CMA is essentially protecting the output of creativity and journalism. What it doesn’t address is the ongoing, foundational scrape for the raw material of AI itself. Google isn’t just using your latest news article to power an AI Overview; it’s using the entire, archived corpus of the public web to build the very architecture of its models. Opting out of the "fine-tuning" is a cosmetic fix when the model was already built on the data in the first place.

This move puts publishers in a stronger negotiating position for licensing deals, yes. We’ll likely see more high-profile content partnerships, where Google pays hefty sums to outlets like News Corp or the New York Times to legally use their content. But that creates a two-tiered internet: the big, established publishers who can afford the legal teams and have the leverage to strike deals, and the vast, independent web of blogs, niche sites, and small creators who lack that clout. For them, the opt-out is a blunt instrument. Do you hide from the AI overlord and risk becoming invisible in its summary, or let it use your work and hope for some trickle-down traffic that likely isn’t coming?

Google’s reaction will be telling. They’ll comply with the letter of the law—adding the necessary switches and robots.txt-like protocols—while aggressively working to make opting out economically irrational. The implicit message to publishers will be: "You can stay out of the AI overview, but you’ll lose visibility in the new default search interface. Your choice." They’ll also double down on securing exclusive, premium content deals to make their AI features better than the opt-out web, further cementing their gatekeeper role.

This UK ruling is a skirmish in a larger conflict: the battle over the data commons. The real, structural fight isn’t about letting an AI summarize your article; it’s about who owns and profits from the collective knowledge and creativity of the internet that was built, in large part, by all of us for free. This regulation is a firewall for content, but the data pipeline remains wide open. It’s a start, but the only thing it truly proves is that the old rules of the internet are dead, and we’re all scrambling to write new ones in the shadow of a few AI giants.

英国竞争与市场出版社协会刚刚向出版商们递上了一把对抗谷歌AI庞然大物的弹弓。从现在起，网站所有者可以阻止其内容被纳入AI概览或用于微调谷歌的模型。这项全球首创的法规，正被竞争与市场协会包装成谈判筹码上的胜利。但请明确：这不是一场革命。而是一场刚刚拉开序幕的战争中，一次战术性的暂停。

英国竞争与市场出版社协会刚刚向出版商们递上了一把对抗谷歌AI庞然大物的弹弓。从现在起，网站所有者可以阻止其内容被纳入AI概览或用于微调谷歌的模型。这项全球首创的法规，正被竞争与市场协会包装成谈判筹码上的胜利。但请明确：这不是一场革命。而是一场刚刚拉开序幕的战争中，一次战术性的暂停。

表面上，这对出版商而言是一次意义重大的胜利——他们一直无奈地看着自己的新闻内容（往往是使谷歌搜索结果有价值的素材）被爬取、摘要，然后装进一个整洁的、吸引广告的框框里返还给用户。赋予他们退出选择权，是基本的数字产权问题。这迫使谷歌承认内容管道并非单向索取的通道。尤其是新闻机构，多年来随着其业务模式在聚合化、平台中介化受众的压力下崩溃，一直对此大声疾呼。

但关键的判断在于：这项法规治标不治本。病根在于像谷歌这样的平台与网络其余部分之间悬殊的数据不对称。通过允许出版商退出AI训练和概览收录，竞争与市场协会本质上是在保护创意和新闻产出的成果。但它没有解决的是，对AI本身原始素材持续进行的、根本性的爬取行为。谷歌不仅使用你的最新新闻文章来驱动AI概览；它是在利用整个公共网络的存档语料库来构建其模型的基本架构。当模型本身就建立在这些数据之上时，选择退出"微调"不过是表面修饰。

此举确实让出版商在授权协议谈判中处于更有利地位。我们很可能会看到更多高调的内容合作，其中谷歌向新闻集团或《纽约时报》等媒体支付高额费用，以合法使用其内容。但不要搞错：这场战争的核心并不是付费墙或版权诉讼。它是关于模型训练数据的采集管道本身。这项法规赋予出版商的能力，本质上是给这根管道安装了一个水龙头——但管道本身依然庞大、复杂，并且很大程度上仍在公众视野之外运行。