Mark Zuckerberg has published a policy document on the future of artificial intelligence, naming not the technology itself but its concentration in the hands of "individual companies or states" as the main enemy. The phrasing is convenient: it allows him to simultaneously appear as a defender of AI democratization and lead a company that has just come under investigation by the European Commission for design that creates addiction in teenagers.
Openness as a business model, not charity
Meta is resuming publication of open models — Muse Glimmer and later Muse Spark 1.2. This is not a gesture of goodwill but a direct counterattack against OpenAI and Google, whose models are closed. If thousands of developers worldwide build products based on Meta's models, the company gains an ecosystem of influence without needing to control each application directly. This is the practical answer to the question "who benefits from openness" — not users in the abstract, but Meta itself, which converts open weights into infrastructural dependence of competitors.
Warnings about regulation are about California and Brussels
Zuckerberg's call not to "over-regulate" AI in the United States comes against the backdrop of states already adopting their own laws on model safety, and the European Union pressing through the Digital Services Act. The argument about loss of competitiveness is an industry classic: it works when regulators fear falling behind China or other players. The practical question is: will US lawmakers agree to postpone restrictions knowing that the same Meta is simultaneously suing the EU over design that harms teenagers.
A billion dollars and parental notifications — reputation compensation
A $1 billion fund for communities with Meta data centers and new notifications to parents about teenager conversations with a chatbot about suicide — these are concrete, measurable steps that are easy to present as care. But they appeared simultaneously with the manifesto about AI "freedom," and this synchronization itself is worth attention: the company is simultaneously asking for less regulation and demonstrating that it regulates itself more effectively than the state.
AI centralization is indeed dangerous — but the question is whether a company with a market capitalization of over a trillion dollars resolves this danger by publishing open weights according to its own rules.
If over the coming months Meta receives a weakening of regulatory pressure in the US while simultaneously facing a fine from the European Commission, it will show that Zuckerberg's manifesto worked as a lobbying tool, not a philosophical position.
