Over the past few weeks, Canada’s Senate has been studying Bill C-18, the Online News Act. It’s the Liberal government’s controversial effort to make rules around how Big Tech companies compensate the journalism organizations whose work appears on their platforms.

In the 11 months since the Liberals introduced the bill in April of 2022, however, much of what we thought we understood about digital media, tech platforms and the whole experience of getting information online has been upended by a Silicon Valley artificial intelligence startup.

Since OpenAI launched ChatGPT at the end of November, it has been a sensation. A language-model tool built on generative AI, it is the fastest-growing consumer app ever, with 100 million users just two months after launch. It has passed the bar exam, written code and with the launch of its latest version, ChatGPT-4, can even share recipes based on what’s in your refrigerator. It has also made clear that big questions about the workings of everything from schools to policy are more immediate than many of us assumed.

Those big questions are also facing journalism, and any business built on delivering news and information to online readers.

Before I go any further, let me put my cards on the table: I’m writing this in my capacity as founder, CEO and publisher of The Logic, not as its editor-in-chief. It’s the same capacity in which I testified in September at a House of Commons committee meeting on C-18, which I generally support. Unlike most major Canadian news publishers, and a host of prominent smaller ones, The Logic hasn’t signed any funding deals with Big Tech platforms, and isn’t affiliated with any of the publishers’ lobby groups. Some other independent publishers disagree with me about the bill’s merits.

At The Logic, we’ve taken every step to make sure the work I do on this file as CEO doesn’t compromise the newsroom’s independence. Since September 2020, when the potential journalistic conflict of interest became apparent, I have recused myself from all newsroom conversations on these matters, and have no say in how The Logic’s reporters and editors cover this file.

With that established, let me come back to the profound impact generative AI is already having on the business of digital content.

The ChatGPT interface spits out well-researched, essay-length copy that is as good, if not better, than a high school senior’s final term paper. Unlike search results, which crawl the internet to serve users links that drive them back to the originating source, generative AI platforms like ChatGPT—and Bard, the ChatGPT competitor Google rushed out after declaring a “code red” in December and calling founders Sergey Brin and Larry Page back into the company to help focus its AI efforts—are trained on information harvested from the internet and present definitive answers or essays on a platform. 

Think of it as the difference between searching for an answer to the infamous question, “What time is the Superbowl?” in 2011 versus 2023. Then, a search platform would send you to The Huffington Post for the answer. In 2023, your query produces an answer that keeps you on the search platform, serving up the answer in a stub created from information crawled from publishers’ websites.

Generative AI takes this to the next level, spitting out not just short, functional answers, but full text mined from all the sources on the internet. There is currently no linking the user back to the originating source material, and in some cases these tools scrape data from behind paywalls. (Bing’s version does at least include citations.) 

I asked ChatGPT if its training data includes news sources. The chatbot told me, “Yes, my training data includes various sources of information, including news articles from reputable sources such as The New York Times, BBC and CNN, among others. These sources provide a broad range of current events and topical information from around the world.”

If a search platform can give you as much information as you care to read about the latest developments in Ukraine, say, or bring you up to speed on the latest scoops in the collapse of Silicon Valley Bank, then you won’t need to click a few more links, read a news site or two and maybe help a news publisher make money, either by buying a subscription or being served ads.

Publishers everywhere are waking up to this existential threat. According to Insider, some U.S. publishers expect the AI technology to spark litigation, as they argue that scraping their content violates their terms of service while tech companies could potentially cite the same fair-use doctrine on which media companies have so often relied.

The same Insider report reveals that News Corp, which among many other things publishes The Wall Street Journal, and which has already struck rich deals with Big Tech companies for the use of its work, is talking to “an unnamed tech player” about how media companies can “be compensated for the use of their other content in training AI engines.” 

In Canada, the contentious debate over C-18 now colours almost every aspect of the broader conversation we need to have here about the future of the news business. But south of the border, The New York Times announced early last month it had expanded its existing commercial agreement with Google. A short press release announcing the deal made no mention of the terms of the agreement, but on an earnings call later that week, New York Times Company CEO Meredith Kopit Levien said it would “begin to see the financial benefits of this deal starting in 2023.”

What’s notable about the Times’s deal is that, unlike similar deals between Big Tech companies and Canadian publishers, it was struck without the looming threat of legislation. The U.S. version of Canada’s bill, the Journalism Competition and Preservation Act, died in Congress last December, and while a new version of the legislation may appear, none has yet been introduced.

It shows once again that, whatever the state of things in Canada, where Meta has said it will pull news content from Facebook if C-18 passes and where Google has tested removing news from search results, Big Tech ultimately remains willing to deal with news publishers.

If generative AI is going to reinvent how we get information online, it strengthens both the practical and the moral case for the tech giants to compensate publishers.

Practically speaking, platforms like ChatGPT and Bard will only be as good as the results they create. To be effective, they’ll need to draw from fact-based, accurate journalism.

As for the moral case, if you think it’s sometimes hard now to figure out what’s true online and what isn’t, you haven’t seen anything yet. 

As the MIT Technology Review’s Melissa Heikkilä phrased it, “AI language models are notorious bullshitters, often presenting falsehoods as facts.” That’s without considering how they might be deliberately used to generate falsehoods.

This “bullshit” on a mass scale could, according to Columbia’s Emily Bell, make the existing disinformation crisis seem like child’s play. 

“The real peril lies outside the world of instantaneous deception, which can be easily debunked, and in the area of creating both confusion and exhaustion by ‘flooding the zone’ with material that overwhelms the truth or at least drowns out more balanced perspectives,” Bell wrote.

Put another way, if artificial intelligence chatbots want to write a factual second draft of history, they’ll need human muckrakers—boots-on-the-ground reporters and editors—to write the first. 

This is where C-18 could be helpful to both publishers and platforms. The legislation forces both parties to come to the table for licensing agreements. Big Tech should want these licensing deals because it alleviates the long-term copyright challenges and expensive lawsuits generative AI is sure to spark. If you license the work, then you are free to use it as you see fit for your algorithm. (I haven’t even touched on how third parties like Slack and Salesforce, with access to the AI models’ APIs, are already launching integrations that will leverage publishers’ content.) And, let’s be honest, the cost of these deals would be a fraction of the cost of the computing power needed for these AI platforms in the first place. 

I take no pleasure from the fact that news publishers are in the position of asking tech companies for compensation. There’s no doubt publishers are partly to blame for how we got to this point. We made what seems like hundreds of incremental decisions in an effort to build audiences, without realizing we were giving up our direct relationship with those audiences to tech platforms. Publishers played checkers while the platforms, to their credit, played chess. I wrote more about this here. 

So be it. But if such agreements are going to exist—and Google’s deal with The New York Times wasn’t signed under duress, remember—the legislative framework of C-18 provides an opportunity for all involved in Canada to forget the past and work toward a mutually beneficial future. 

It’s about future-proofing the vital role of journalism—and ensuring journalism businesses old and new can compete with each other on a level playing field—whatever new technologies come next.