◢ AGI Lab Power Maps

Menlo Park · USA · FOUNDER LOCK · DUAL-CLASS

Who controls Meta Superintelligence Labs?

ULTIMATE CONTROL Mark Zuckerberg — majority voting control of all of Meta

A 'flat' research org sits beneath a company one person controls outright through 10-vote Class B shares.

74/100concentration of ultimate control, the essays' editorial reading (0 = structurally checked, 100 = one person decides). How to read it

Chain of command

  1. THE TOOLMeta Superintelligence Labs

    Co-led by Alexandr Wang & Nat Friedman; 'TBD Lab' core team. Built via a $14.3B Scale AI deal + aggressive talent raids.

  2. OWNS ITMeta Platforms, Inc.

    Public company; the lab is a division with no separate governance.

  3. CONTROLS METAMark Zuckerberg

    Owns ~99.7% of Class B shares (10 votes each) → ~61% of total voting power on ~13% economic stake. Cannot be outvoted.

  4. SAFETY CHECKNone binding

    No independent trust or external veto; deployment decisions are corporate, ultimately the founder's.

  5. OPEN QUESTIONWho can say no?

    Structurally, no shareholder coalition or board can override Zuckerberg.

At a glance

ModelLlama series + 'superintelligence' push
FormDivision of Meta Platforms, Inc.
EquityZuckerberg ~13% economic / ~61% voting power
Workforce~3,400 staff [10]

Ultimate control: Mark Zuckerberg. ~61% of Meta's votes on ~13% economics. Cannot be outvoted. No independent veto. [6]

~13% → 61% Mark Zuckerberg's economic stake vs. his voting power, via 10-vote Class B shares. He cannot be outvoted. [6]

The pitch and the record

They say: “Safety has a real seat at the table here.”

At Google DeepMind the AGI Safety Council and RSC are advisory ; final authority rolls up to Alphabet, where founders Page & Brin hold super-voting control. At Meta and xAI there is no independent safety veto at all. [9] [6]

The essay's proposed fixes

A prescription from "The People Inside", not a fact about Meta Superintelligence Labs and not something the lab has agreed to.

END THE FOUNDER LOCK One person holds ~61% of Meta's votes on ~13% economics and cannot be outvoted. There is no independent safety veto. Superintelligence cannot be governed by a single uncheckable will.

  1. Strip super-voting power over AI governance. Zuckerberg's 10-vote Class B shares must not control superintelligence decisions; create a separate governance class for the AI mission.
  2. Seat an independent board majority with a safety veto. Directors not appointed by or loyal to the founder, holding a real deployment kill-switch.
  3. Put the weights under an independent mission trust. Owned by the trust, licensed to Meta on revocable terms tied to the public-benefit mandate.
  4. Cap returns and lock the public benefit. Bind the AI division to a return ceiling and an enforceable benefit charter, not Meta's at-will profit motive.

Sources the essays cite for Meta Superintelligence Labs

  1. [6] Meta dual-class structure — Zuckerberg ~13% economic / ~61% voting. citizen.org · ainvest.com
  2. [9] Google DeepMind safety governance — AGI Safety Council & Responsibility & Safety Council (advisory). deepmind.google/responsibility-and-safety
  3. [10] Approximate workforce figures: OpenAI ~4,500→8,000 (2026); Anthropic ~3,000–5,000; DeepMind ~6,000; Meta Superintelligence Labs ~3,400; xAI ~1,200. saastr.com · engadget.com · builtin.com · time.com

Read the full essays: Who Would Own God? · The People Inside.