At the start of this year, the UK’s Competition and Markets Authority (CMA) proposed a series of changes to how Google’s search service operates in the UK, including allowing publishers to opt out of their content being used either as cited sources within Google’s AI Overviews feature, or for wider training of the tech giant’s AI models. The proposal came in response to complaints from media companies that Google has been leveraging its market power in search to impose unfair conditions on publishers, essentially strong-arming them into allowing the tech company to feed their content into its AI systems.
Now, following a consultation period, Google has agreed to play ball. While it continues to dispute claims that the introduction of AI features into its search interface is harming publishers, the company says it is developing controls “to let sites specifically opt out of generative AI features in search”.
At a glance, that looks like a win for publishers. But the devil is in the details, and submissions to the CMA’s consultation from media groups and Google itself show significantly different visions for how these controls should operate.
Fine print and fine tuning
The CMA’s consultation drew responses from across the publisher landscape, with the BBC, Guardian Media Group, Financial Times, Independent Media, and DMG Media all submitting responses, alongside industry trade groups including the AOP, News Media Association, and Publishers Association.
These responses all welcomed the CMA’s proposal to introduce greater controls for publishers to manage how Google’s AI systems are able to interact with their content. But they warned that if not designed correctly, these controls could end up being fairly ineffective.
Publisher respondents were generally in favour of the CMA forcing Google to separate out its AI crawler from its regular search crawler, making it easier for them to monitor and manage how and when their content is being scraped. Google has argued that this would be expensive and slow to implement — a view which the CMA seems to sympathise with. While several publishers accepted the CMA’s position in their responses, they continued to argue that enforced crawler separation would give media companies a stronger hand in their negotiations with Google.
The BBC, for example, said that crawler separation would give publishers “at least the opportunity to monitor the activity of Google’s various crawlers, rather than rely solely on Google implementing separation ‘behind the scenes'”. This detection ability would make it easier to understand how often their content is being used by Google’s AI tools, and therefore better gauge its value.
Another common ask was that publishers are able to opt in and out of distinct AI use cases, rather than them all being bundled into one consent mechanism. So while publishers want a specific opt-in feature for search, they’d also like to be able to opt in to the different ways their content might be used by AI tools, such as training, grounding, and fine-tuning.
“Publishers must be able to exercise distinct, granular controls over different uses of IP, including grounding, training, and fine-tuning,” the Financial Times said in its response. “As noted in our previous response to the relevant CMA RFI, we believe it essential that we are able to do so at both directory and page level across all use cases. Bundled or imprecise controls would significantly weaken our autonomy and ability to negotiate effectively, undermining the CR’s objectives.”
There are also concerns about whether there will be sufficient protection to prevent Google from getting access to publisher content via deals with third-party sources. Trade group News Media Europe suggested that Google “should be required to remove and exclude that content from any datasets it acquires, regardless of source, prior to any AI use,” adding that “otherwise, publisher choice risks becoming illusory”.
And of course, as is always the case with any sort of consent mechanism, there’s the debate over whether publishers should have to opt out of or opt into Google’s AI products. DMG Media strongly supports the latter.
“Publishers should not be required to take active steps to prevent their content being used in Google’s AI services,” the media group said in its submission. “Instead, publishers should be opted out by default […] Publishers are under comparable obligations regarding user consent under the GDPR, and this approach would be consistent with those principles. Having opt-out as the default would also strengthen the bargaining power of publishers by reducing the risk that content is used by default or through inadvertent consent, and by ensuring use occurs only following an active decision to opt in.”
Follow VideoWeek on LinkedIn.
