AI safety meets public ownership: who gets the board seats?
El-Sayed restated his public-governance proposal on CBS on September13. Interview. The underlying documents below are June proposals, newly added to this investigation on September14.
The debate includes an explicit bid to change who owns and governs AI businesses. El-Sayed’s platform would require public benefit charters and majority public board representation. It also proposes independent safety assessment, restrictions on certain applications and controls over access to computing resources.
The campaign announcement connects that agenda to Sanders’s ownership plan. Its stated justification combines public contribution to AI, worker protection and resistance to concentrated private power. That is the campaign’s case for the proposed redistribution of decision rights.
Sanders’s office summary describes a public fund with a commission managing its holdings. The office’s headline valuation is an estimate of what the proposal might produce, not money already collected. The linked text is a legislative draft; this update establishes no passage or enacted transfer.
Two powers that should not be confused
Ownership provides economic claims and the rights attached to shares. A guaranteed majority of board seats is a separate corporate-governance rule. Neither should be casually translated into unlimited control. Board composition, voting classes, conflicts rules and the applicable mandate determine what a representative can actually decide.
The institutional contrast with embedded laboratory evaluators is useful. An evaluator observes, tests and reports under an access arrangement. A shareholder representative participates in company governance. A regulator may compel compliance. Calling all three oversight conceals the practical question this investigation asks: which decisions move, to whom, and with what means of challenge?
What would change the picture next
Follow formal legislative records, amendments and the selection rules for any proposed public representatives. For the campaign platform, an implementing text would show how election and appointment alternatives are meant to work. For the existing evaluator proposals, contracts remain the missing counterpart.
The campaign also explicitly prohibits mass political surveillance, AI targeting of dissent and warrantless surveillance. Those limits belong in the account alongside its proposed controls.