Meta is walking away from the open-source AI playbook that made it famous in artificial intelligence circles, pivoting instead toward a proprietary model developed in partnership with Alexandr Wang, the entrepreneur behind Scale AI.
According to Startup Fortune, Meta has said outright that its old open AI strategy "no longer works" — a striking admission from a company that built much of its AI credibility by publicly releasing models like Llama for anyone to download and use.
The shift has been in motion for roughly a year. According to CNBC, Meta tapped Wang about a year ago to help build a new AI model — and now Zuckerberg faces the task of convincing the world the change in direction is the right one.
Wang is a prominent figure in AI infrastructure: Scale AI supplies the data labeling and training pipelines that underpin many of the industry's most powerful models. His involvement signals that Meta is serious about building something more tightly controlled and, potentially, more commercially competitive.
The reversal is significant because Meta's open approach was both a business strategy and a philosophical stance. By releasing powerful models freely, Meta positioned itself as a counterweight to closed competitors like OpenAI and Google — and won goodwill from researchers and developers worldwide. Abandoning that stance means Meta is betting that proprietary control will deliver better returns, even if it costs the company some of that community trust.
If Meta's open AI era truly is ending, it removes one of the most powerful forces pushing the industry toward openness — and could tilt the competitive landscape further toward a handful of companies controlling the most capable AI systems.