Let’s Check the Receipts
Mark Zuckerberg spent part of this week telling reporters that the real threat to AI’s future isn’t Meta, it’s the labs being too careful. In a New York Times interview, he took aim at OpenAI and Anthropic, arguing that their safety focused approach amounts to concentrating power in a few hands, and he called the discourse coming out of those labs “filled with doom.” Openness, in his framing, is the responsible choice. Meta is the company handing power back to regular people.
Every tech giant has a founding myth. Meta actually has two, and they don’t agree with each other. One says Meta believes in putting power in people’s hands. It’s worth checking that against the platform that made Zuckerberg a household name in the first place.
Because Facebook, the product, has never been open. Not by accident, and not for lack of trying to close it further. This is Meta’s open platform history, and it doesn’t match this week’s talking points.
The Social Graph Was the Product, Not a Public Good
In 2010, Facebook launched Open Graph, an API that let outside developers build apps pulling in a user’s data, and their friends’ data, with a couple of taps of “Allow.” It sounds generous. It also created exactly the conditions for what happened next: a Cambridge academic built a quiz app in 2013 that harvested profiles far beyond the people who actually took the quiz, then handed that data to a political consulting firm called Cambridge Analytica. By the time the story broke in 2018, up to 87 million people’s information had been swept up, mostly in the US, and Facebook was facing a $5 billion FTC fine, the largest privacy penalty in US history at the time.
What Facebook did next is the part I find most telling. It didn’t hand users more structural control. It locked the doors: shutting down broad friend data access by 2015, then stripping down the Events, Groups, Pages, and Instagram Platform APIs in April 2018. Good for privacy. Also exactly the move a company makes when it wants total control over its own ecosystem.
Openness, Until It Threatens You
The most revealing example predates Cambridge Analytica by five years, and it has nothing to do with privacy. In 2013, Twitter launched Vine, a short video app that needed access to a user’s Facebook friends to work properly. According to the FTC’s 2020 complaint, Facebook shut off that API access specifically to blunt Vine as a competitive threat. Zuckerberg’s own internal correspondence, cited in the case, put the philosophy plainly: it’s better to buy than compete. That’s exactly what happened next, with Facebook acquiring Instagram in 2012 and WhatsApp in 2014, both of which the FTC later argued were bought to neutralize them before they became real rivals.
Where the Antitrust Story Actually Stands
The FTC sued Meta for illegal monopolization in December 2020. The case took the long road: dismissed, refiled, tried. As of this writing, that outcome is still genuinely undecided, so I’d rather say that plainly than pretend it’s already settled.
What isn’t in dispute is the underlying conduct: the API cutoffs, the “buy don’t compete” emails, years of walling off the social graph from anyone who might use it against Facebook’s interests. Whether that adds up to an illegal monopoly is a legal question the courts are still deciding. Whether it adds up to a company that treats openness as a liability rather than a value is a much easier one.
The Irony Doesn’t Write Itself
So here’s a company with a well documented history of building closed systems and defending them hard, this week positioning itself as the last honest defender of open AI. Maybe that’s a genuine change of heart. Maybe it’s just good timing. Both things can be true at once, which is sort of the whole point of a founding myth.
What I’ll say in Meta’s favor, and I mean this: on the AI side specifically, its open source record is more substantial than people give it credit for. That’s a different story with different receipts, and I’ll get into it next time.
