Mark Zuckerberg

On 10 August 2026 Mark Zuckerberg published "The Future is for Everyone" on meta.com, some 6,500 words signed only with his first name. The argument cuts against the industry's default framing of alignment: because "humanity is not a monoculture", people's values encode opposing trade-offs, and no technical solution can align with contrary interests at once. Hence the claim that there is no singular benevolent superintelligence. Safety becomes a question of how power is distributed, and the most dangerous scenario he names is frontier labs training the strongest models and keeping them to themselves.

The commitments attached are concrete: a personal agent with a fully private mode where "even Meta" cannot grant access, free versions for billions alongside a dynamic auction for paid compute, intermediate training checkpoints handed to government instead of a blocking end-of-cycle review, and a resumption of publishing some open source models. Meta's independent board is to approve safety release criteria, and the text concedes the company remains founder-controlled.

Two weaknesses sit inside the manifesto. Sourcing is close to absent: the employment statistics, the HuggingFace incident and the Chinese nuclear capacity figures go unreferenced. And alignment ends up as a consequence of adoption, since billions of people using and scrutinizing personal agents is offered as proof the problem was solved.

Meta's acquisition of Manus in December 2025 for more than $2 billion, its third largest, had already put IA agentique at the centre of the same programme.