OpenAI cuts ties with Scale AI after Meta deal

In the cutthroat AI race, there is one rule everyone understands: you do not let your key data provider climb into bed with your fiercest competitor.

On Wednesday, an OpenAI spokesperson confirmed to Bloomberg that the company is officially winding down its work with Scale AI and cutting ties completely. The move marks a sharp reversal from just days ago, when OpenAI CFO Sarah Friar hinted that the partnership would continue.

OpenAI claims the breakup was already in motion before Meta announced its multi-billion-dollar investment in Scale AI and hired its CEO Alexandr Wang. An OpenAI spokesperson stated that the lab had already been shopping around for more specialized data vendors to fuel its frontier models. But the timing leaves little doubt that Meta's aggressive power play accelerated the exit.

The fallout of losing Switzerland

Scale AI spent years branding itself as the neutral picks-and-shovels provider for the entire AI industry. That illusion has now evaporated.

OpenAI is not the only giant heading for the door. Reuters reported last week that Google is also weighing plans to drop Scale AI as a data provider. Meanwhile, rival labeling startups report a surge of inbound interest from AI developers hunting for genuinely independent partners.

Scale AI scrambled into damage-control mode on Wednesday. In a blog post, the company's general counsel insisted Meta would receive zero preferential treatment, pledged that customer confidential data would never cross borders, and noted that Wang would no longer oversee daily operations.

Yet the damage to its core data labeling engine appears irreversible. In a separate post the same day, interim CEO Jason Droege announced the startup will double down on its applications arm, shifting focus to custom AI products for enterprise and government clients.

Why it matters

When training frontier frontier models, training recipes and RLHF annotation data are trade secrets of the highest order. Having your data supplier's founder work directly for Meta is a non-starter for any serious AI lab. If you are building models and relying heavily on a single external labeling vendor, this is a clear reminder of how brittle data supply chains can be when platform politics intervene.

For the AI ecosystem, this marks the end of Scale AI as the undisputed data backbone of the industry. Meta scored a massive talent and infrastructure win by bringing Wang into its tent, but in doing so, it effectively poisoned Scale AI's relationships with the rest of Big Tech.

We think OpenAI and Google made the only logical call here. No legal firewall or corporate pledge can offset the paranoia of your proprietary training pipelines being compromised. Scale AI's pivot toward government contracts and enterprise software might keep the lights on, but its original dream of being the universal data utility for frontier AI is officially over.