Google–California Deal Falls Short Where a Data Tax Would Succeed

Google’s agreement with California to commit over $172.5 million to journalism and artificial intelligence is an ad-hoc way to sidestep tax and regulatory policies in favor of voluntary, tax-deductible contributions to nonprofits.

To safeguard the future of journalism, lawmakers should avoid short-term agreements, “link tax” saber rattling, and corporate payouts in favor of broader taxation of companies such as Alphabet Inc.’s Google. A tax on user data used for advertising purposes and third-party data used for AI model training, along with revenue allocation to targeted initiatives, could provide ongoing support—rather than temporary infusions of capital—for journalism. Such a strategy would better align tax policy with the outsize role tech companies play in California.

California also committed $70 million to journalism initiatives as part of the deal that supplanted more complex regulatory and “link tax” legislation—proposals that Google opposed.

The agreement, negotiated between Google and California Assemblymember Buffy Wicks (D), who introduced legislation requiring online platforms to pay for access to digital journalism providers, may be the first of many. This could provide a problematic model for other tech giants and other states.

The California Journalism Preservation Act would have required tech companies pay for access to journalism platforms and provided protections from retaliation by tech companies for journalism platforms that opt not to provide access. A similarly targeted bill, which remains in committee, would amend state law to tax companies that engage in “data extraction transactions”—sale or use of user data for advertising purposes. If passed, those tax proceeds would support journalism through grants and fellowships.

The deal struck between Google and California is part of a wider trend where tech giants negotiate individual agreements to avoid more stringent regulations or taxes. Google made a similar move in Canada, where it brokered a deal to pay $73.6 million annually to newsrooms to avoid the country’s Online News Act, which would have required tech companies to negotiate with online news organizations to use or link to their content.

Facebook parent company Meta Platforms Inc. forged similar deals in Australia before resisting additional measures there and ultimately opposing the Canadian “link tax” and any payouts to avoid its passage.

These deals reflect a systemic reluctance to use tax systems to support journalism and hold tech companies accountable. A comprehensive “data tax” would be more effective and equitable than ad-hoc arrangements between states and tech giants. This would target two primary sources of revenue for tech giants: user data mined for advertising purposes and external data sources used for training AI models and content aggregation.

Today, the technology industry’s appetite for data has only become more ravenous with the advent of large language models such as OpenAI’s GPT-4 and Google’s Bard. Companies such as Google and Meta generate immense profits from advertising, which in part necessitates mining user data for targeting purposes and, increasingly, by ingesting content from user and third-party sources to train their AI models.

Even considering the shelved legislation, this Overton window fails to adequately compensate originators of data whether they’re users or news outlets. A data tax with components for levying taxes on companies based on revenue generated from digital advertising and data ingestion would be a better fit.

The portion of a tax aimed at compensating users for their data can be keyed to advertising revenue, and taxes on AI models levied based on elements of the language model itself—such as parameter size, which is a rough approximation of the complexity of a given model. Complexity generally reflects the amount of training data a model has been exposed to. This means a tax on parameters is something of a proxy for a tax on data ingested.

Such a policy would allow equitable and progressive rates of taxation, revenue distribution, and a reduction of power imbalances between states and large tech companies. Portions of its revenue could be allocated to state initiatives supporting journalism and AI development. The tax would need clear definitions of what constitutes user data and content ingestion, as well as enforcement and compliance mechanisms, including penalties for underreporting or evasion.

Despite these policy drafting difficulties, the agreement between Google and California is, at best, a short-term boon to journalism—its long-term effectiveness is in doubt, particularly with the five-year deal term. Even in the near future, Google’s actual out-of-pocket dollar contribution is an open question. The tax deductibility of Google’s contributions could undermine their overall benefits to tax revenue.

Of the $172.5 million pledged, most will be tax-deductible if the contributions are made to nonprofits. This would substantially reduce the positive tax revenue impact, effectively using the fungibility of money to borrow from income tax receipts to subsidize Google. Taxpayers may wind up funding part—perhaps much—of Google’s contribution.

Over the agreement’s five-year period, the state’s $70 million commitment to journalism and AI initiatives heaps more public funding on top of the subsidies enjoyed by Google—a boost to journalism, to be sure, but not one financed by tech. Ultimately, this means a significant portion of the funds would come from taxpayers.

The state deserves a more equitable policy that could support journalism and other digital platforms more permanently.

Andrew Leahey is a tax and technology attorney, principal at Hunter Creek Consulting, and adjunct professor at Drexel Kline School of Law. Follow him on Mastodon at @andrew@esq.social

Read More Technically Speaking