Google has said it is exploring new “value exchange models” with publishers and accepted the need for independent regulation of AI.

The tech giant’s paper, A Pragmatic Approach to AI Governance in America, has defended the use of publicly available online content for training AI models, but said it is “piloting novel ways to partner with websites whose content meaningfully contributes to the freshness and factuality of generative AI responses”.

It also said, while training on training AI models on “publicly available web data” should remain legal, AI model developers should give website owners choice and control over whether their content is used in model development using robots.txt tags.

“While this use is legally protected, responsible model developers should give website owners choice and control over whether their content is used in model development via simple, machine-readable robots.txt tags, like the Google-Extended control. Such controls enable markets to work and allocate value where appropriate.”

A leading industry executive recently told Press Gazette that he has seen instances of AI companies not respecting robots.txt signals and scraping his site more than 12 million times in the past three months.

Google supports commercial licensing deals

AI developers should explore partnerships with content owners, the policy paper states, which is what Google claims to be doing: “Like other platforms and AI model developers, Google is exploring new types of partnership and value-exchange models,” it said.

“For instance, we are piloting novel ways to partner with websites whose content meaningfully contributes to the freshness and factuality of generative AI responses through the process of grounding. We have also entered into deals in which we are paying for access to and delivery of diverse types of specialised, non-public content, including creative and educational content.”

Google added that collaborations with publishers of “specialised, non-public” content “are the ones most likely to be sustainable for developers and meaningful for the ecosystem”.

Proposal of independent regulation body

To protect US “innovation” and lead in AI, a unified framework is needed, Google argues, proposing an independent, regulatory organisation “that can keep pace with fast-moving AI research and development”.

“A federally overseen frontier AI regulatory organisation [FARO], coupled with targeted policies for widely deployed AI models, can address both national security and consumer protection risks while promoting economic, scientific, and social progress,” it said.

“Achieving this vision depends on proactive, evidence-based collaboration among policymakers, industry leaders, and civil society.”

It added that a FARO could “standardise and verify” that AI companies meet safety and security standards, publish AI safety frameworks and complete annual government audits before releasing models to market.

Barry Adams, founder of the News & Editorial SEO Summit, told Press Gazette the proposal of an independent body “means they can get ahead of the legislative curve and influence the inevitable regulation coming for the AI industry, and ensure they secure a decent deal”.

Google has faced increasing regulatory scrutiny in recent months: earlier this month, in a world first, UK regulators told Google to give publishers control over how their content is surfaced in AI answers.

Adams added: “Unsurprisingly, Google appears to promote a fairly soft-touch approach to AI regulation, especially for already widely used AI applications (such as their own search engine). At least they recognise the need for regulation, but it would be folly to follow their recommendations to the letter, as only Google would benefit.

“An independent US body would also help in other markets, such as the EU, which has been much more restrictive in its approach. A US-based oversight structure would serve as an example for the EU and other markets, and perhaps ensure that collective organisations such as SPUR aren’t easily able to extract maximum value from the AI-content exchange.”

Steve Wilson-Beales, SEO and content strategy consultant, said: “Publishers and copyright-holders will be disappointed, as it will now be on them to opt out of AI activity.”

SPUR (the Standards for Publisher Usage Rights coalition) is an AI news licensing standards coalition made up of major publishing members including The Guardian, Financial Times, Telegraph, BBC and Sky News.

[Read more: Publishers to bill AI firms for unwanted scraping – and take them to court if they don’t pay]

Copyright concerns: focus on output not training

On copyright concerns, Google said focus should be on whether content “actually copies” existing work, rather than whether the content was used for AI model training.

“The focus should again be on outputs – in this case, whether a specific image or piece of text actually copies an existing work, regardless of how it was created.”

It added that “technical safeguards” can help prevent AI output reproducing copyrighted content, such as the reporting and taking down of infringing content.

Google also said it supports regulation against deepfakes, protecting individuals from AI-generated replicas of their voice and likeness, and highlighted Generative AI as a “powerful tool” for human creativity.

Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our “Letters Page” blog