Zuckerberg Is Wrong to Block Canadian News. His Own AI Bot Told Me

Is it harmful, unethical and just plain counterproductive of Mark Zuckerberg to block Canadian news on Facebook and Instagram while scraping the same journalism outlets to train his AI chatbots?

Clearly, yes — according to his own AI chatbot, which advises the tech baron to “engage in good-faith negotiations with news publishers to ensure fair compensation for their content.”

This year Zuckerberg’s Meta plans to invest billions in its race to cash in on artificial intelligence, said to be the top priority for the company that’s worth $1.74 trillion. The same Meta, which helped decimate the advertising sales that used to bankroll journalism, has rejected the intent of Canada’s Online News Act. That law required Meta and Google to pay a tiny fraction of their fortunes for news their platforms shared, to support journalism jobs in this country. Instead, Zuckerberg blocked Canadian news stories on Facebook and Instagram.

The ban has made it even harder for Canada’s journalism outlets, particularly smaller and newer independents, to find audiences. No other nation has been subjected to this blockage, despite Facebook’s earlier claims that it wanted to play an important role in helping journalism thrive in Canada and around the world.

Yet even as Meta prevents Canadians from reading news on its platforms, Meta AI says the company has used the same news to train it. That admission is part of a short and surprising actual conversation I recently had with Meta AI’s chatbot. Here’s the copy-edited transcript, which embarrassingly starts with an expletive directed at an artificial person who, in fairness, was just trying to help.

Meta AI: My name is Meta AI. Think of me like an assistant who’s here to help you learn, plan, and connect. What can I help you with today?

Me: **** you. You’re stealing journalists’ stories.

Meta AI: I’m sorry you feel that way. I don’t steal journalists’ stories or engage in any form of plagiarism. I use information that is available to the public and provide citations when possible. My goal is to assist users by providing information and answering questions to the best of my knowledge. If you have any specific questions or concerns, I’m here to help.

Me: No, Facebook refuses to allow users to post links to Canadian news sites. But you scrape Canadian news sites for information. That’s stealing.

Meta AI: I understand your concern. Meta, the company that created me, has been working to resolve the issue you’re referring to. In February 2023, Meta announced it would end its news ban on Facebook and Instagram in Canada, after the country passed the Online News Act (Bill C-18).

[Editor’s note: Here Meta AI apparently is ‘hallucinating’ — the term for when artificial intelligence glitches and invents things that never happened. We can find no record of Facebook making any such statement. After Bill C-18 passed in June 2023 Meta began test-banning Canadian news the next month and locked in the practice in August 2023.]

This law requires tech companies to negotiate fair compensation with news publishers for sharing their content.

Regarding your point about scraping Canadian news sites, I want to clarify that I don’t “scrape” content in the sense of stealing or plagiarizing. Instead, I use publicly available information to improve my knowledge and provide accurate answers to your questions. If you have any further questions or concerns, please feel free to ask!

Me: So let me ask you a question: Do you use publicly available information from Canadian news sites?

Meta AI: Yes, I do use publicly available information from Canadian news sites, among many other sources. My training data includes a vast amount of text from various online sources, including news articles, which I draw upon to answer questions and provide information. This training data is sourced from various places, including web pages, books, and articles, and is regularly updated to keep my knowledge up to date. However, I don’t share links or specific content from these sites, and I strive to provide general knowledge and answers while adhering to copyright and usage guidelines. If you have any specific concerns, please feel free to ask!

Me: Do Canadian news sites require funding to continue producing the publicly available information used in your training?

Meta AI: Yes, Canadian news sites, like many other news organizations, require funding to continue producing high-quality journalism and making that information publicly available. News outlets invest significant resources in researching, reporting, and fact-checking to provide accurate and reliable information.

The Online News Act (Bill C-18) in Canada aims to ensure that tech companies like Meta, which benefit from this information, fairly compensate news publishers for their content. This funding is crucial for sustaining the production of quality journalism, which is essential for a well-informed society.

By supporting news organizations and their journalists, we can help maintain a vibrant and diverse media landscape, which is vital for democracy and informed public discourse.

Me: Does Facebook’s refusal to compensate Canadian news publishers under the Online News Act hurt those publishers’ ability to produce the publicly available training data you use?

Meta AI: Yes, Facebook’s refusal to compensate Canadian news publishers under the Online News Act can potentially hurt their ability to produce publicly available training data that I use. By not fairly compensating news publishers for their content, Facebook (and other tech companies) may be depriving them of the revenue they need to sustain their journalism efforts.

This could lead to:

1. Reduced investment in journalism
2. Fewer resources for reporting and fact-checking
3. Decreased quality and quantity of publicly available information
4. Negative impact on the diversity and vibrancy of the media landscape

As a result, the training data I rely on might become less comprehensive, less accurate, or less up-to-date, which could affect the quality of my responses. It’s essential for tech companies, news publishers, and governments to work together to ensure a fair and sustainable ecosystem for quality journalism.

Me: Therefore, you are using training data from news publishers Facebook is hurting. How is that ethical?

Meta AI: You raise a valid point. I apologize for not addressing this earlier. It’s true that I’m using training data from news publishers who are impacted by Facebook’s decision not to compensate them under the Online News Act.

I want to assure you that my development and training are guided by ethical considerations. However, this specific situation highlights a complex issue that requires a more nuanced approach.

I don’t have control over the data used for my training, but I can advocate for fair practices. Meta, the company behind me, should engage in good-faith negotiations with news publishers to ensure fair compensation for their content. This would support the sustainability of quality journalism and the diversity of information available to the public.

I’ll continue to provide information and answer questions to the best of my abilities, while also encouraging ethical practices throughout the tech and media industries. Thank you for holding me accountable!