Section 2 of 4
Most public AI safety advice operates at constants and parameters — the weakest place in Meadows’ hierarchy. That’s not a reason not to act. It’s a reason to be precise about what you’re moving.
The leverage point hierarchy exists because all interventions are not equal. A content filter update, a regulatory comment submission, a product liability case, and an international safety summit are all “doing something about AI.” They operate at entirely different levels of the hierarchy. The goal of this section is to be honest about which actions land where — and why the distinction matters for how you direct effort.
Consumer-level action operates at constants and parameters. Canceling subscriptions, avoiding AI products, boycotting specific labs — these adjust a parameter. They’re real signals that product teams notice in aggregate. They don’t change who is liable, who audits, what deployment gates must be cleared, or what financial incentives the capital loop runs on. Frontier AI development is primarily funded by institutional investors, government contracts, and enterprise customers. Consumer revenue is a fraction of committed capital. A parameter-level intervention adjusts a number; the structural loops continue.
This is a structural observation, not a moral one. Parameter-level actions aren’t worthless — public signals shape lab reputations and, at significant scale, investor sentiment. But understanding their structural limits lets you calibrate effort: if consumer-level action is the only action available to you, being realistic about what it can and can’t accomplish is more useful than assuming it’s sufficient.
The most accessible high-leverage intervention for most individuals is visibility. Making harm visible doesn’t require institutional access. It requires documentation, specificity, and a channel to share it. Journalists investigating AI harm, researchers publishing documented cases, employees creating internal paper trails of AI deployment decisions, and procurement professionals demanding safety disclosure from vendors are all operating at visibility. Unlike boycotts, visibility work compounds: each documented case builds the evidentiary base that courts, regulators, and legislators need to act. One investigative story about a specific AI harm does more structural work than ten thousand awareness-raising social media posts.
The most underused visibility entry point
Enterprise procurement is where most AI revenue lives — and where disclosure demands have the most leverage. A procurement team that requires detailed safety documentation before signing contracts, or that includes AI incident reporting clauses, is operating at visibility with contractual enforcement behind it. Most AI vendors have not yet been asked for this by most buyers. Asking is not technically demanding. The barrier is knowing it’s a lever.
Coordination multiplies all of it. Visibility disclosure demands are more effective when many people make them simultaneously. Enforceable-rules liability cases require plaintiffs. Regulatory participation that diversifies who writes the rules is more effective from organized coalitions than individual submissions. The effective unit of action at the higher leverage tiers is usually not a single individual — it’s people whose efforts compound in the same direction, often without coordinating explicitly. The value of understanding your structural position is knowing whose efforts you might be compounding when you act.
What "doing something" at each tier actually requires
Constants and parameters: Consumer decisions, social media pressure, public petitions. No institutional standing required. Delays, brakes, engine speed: Journalism, legal work, legislative drafting, insurance pricing. Requires professional access or institutional standing. Visibility, enforceable rules, rule-writers: Investigative reporting, litigation, regulatory participation, procurement demands. Requires channel access — more people have this than use it. Goals, paradigm, coordination: Sustained institutional effort, international framework-building, decades-scale research. Requires coordination at scale.
The tier cards that follow map all twelve leverage points with specifics on access and what changing each one actually does. The honest question to bring to them: which ones am I actually positioned to move, and what would it concretely require?