Mark Zuckerberg's case is that Meta can put advanced AI in billions of hands, and on distribution he is plainly right: Meta's apps averaged 3.56 billion daily users in March 2026, so a capability arrives as an app update. Digital Trends has the argument as he made it.
What the pitch leaves out is who decides anything. Users receiving a model do not choose how it was trained or which safeguards it carries, and Meta is spending $125 to $145 billion on the data centres where those choices get made. Forbes sets it against the same day's other news, when more than 1,000 workers at OpenAI, Anthropic and Google DeepMind asked the US government for the means to slow AI down. Access for everyone and control for one company are not the same promise.