Section 4 of 4
The tiers that follow are ranked by structural power. They’re not ranked by who has access to them. This section maps the honest relationship between standing and which levers you can actually move.
The twelve leverage points in the tiers below are not equally accessible to everyone. Access depends on three things that vary by person: institutional standing (what organizations you’re part of and what authority they carry), professional position (what your work gives you proximity to), and civic standing (what legal and political systems you have access to and in what capacity). Understanding which tier you’re actually positioned to move is more useful than trying to act on all of them simultaneously.
Tier 1 — accessible to nearly everyone, limited ceiling
Most public AI safety activity clusters in Tier 1 — not because it’s lazy, but because it’s genuinely accessible without institutional standing. Consumer pressure, public statements, and awareness work are real inputs to the system. Their structural limits are covered in section 2. If your current position doesn’t give access to higher tiers, Tier 1 work that builds documentation and evidentiary record is not wasted — it feeds directly into higher tiers that others will operate. The most valuable Tier 1 contribution is documentation that is specific, attributable, and persistent: not “AI is producing harm” but “this specific system, deployed in this context, produced this specific outcome, on this date.”
Tier 2 — requires institutional access or legal standing
Shortening delays requires journalists and researchers with resources to document harm systematically over time, and regulatory staff with authority to mandate incident reporting. Strengthening brakes requires plaintiffs, plaintiff attorneys, insurance actuaries, and appellate judges. Slowing the capital loop’s acceleration requires legislators and finance officials. If your professional role gives you proximity to any of these — legal work, journalism, insurance, policy analysis, legislative staff — you may have Tier 2 access that isn’t being used. The barrier is usually not standing; it’s the framing of AI as someone else’s domain.
Tier 3 — the most underused tier relative to access
This is where individuals and small organizations can create structural change in the next two to three years, and where the gap between available access and deployed effort is largest. It requires channel access — to investigative journalism, court proceedings, regulatory comment processes, or workplaces deploying AI — rather than institutional authority.
The most underused entry points:
Disclosure demands in enterprise procurement. Most AI vendors have not yet been required to provide meaningful safety documentation as a condition of sale. A procurement team that requires it creates a financial incentive for every vendor who wants that contract. This is visibility work with contractual force behind it.
EU AI Act regulatory consultations. Open to non-EU organizations in most cases. The EU AI Office is actively developing the technical standards that will determine what “transparency” and “safety” mean in enforceable law. Public consultation periods are undersubscribed relative to their structural significance.
Workplace AI policy participation. Every large employer deploying AI is making decisions about what documentation it keeps, what disclosure it provides to workers, and what consultation it conducts. These decisions are currently being made without meaningful participation by most people they affect. The EU AI Act’s Article 26 requirements make this a legal mandate in EU jurisdictions; in the US, it requires organized professional pressure.
Legal standing as a plaintiff or witness. The liability cases accumulating in US federal courts (section 7 of Part 1) need plaintiffs and documented harm. Individual experience of AI-caused harm that is documented, attributable, and pursued is structural input to enforceable rules — the lever with the fastest activation potential once an appellate court sets precedent.
Tier 4 — the longest timeline, the highest ceiling
Changing the system’s goals, dislodging its operating assumptions, and coordinating competing actors are not individual actions. They are outcomes of many actors working in the same direction over years. Individual contributions at this tier: funding interpretability research that makes AI risk legible and verifiable, publishing or amplifying work that shifts the operating assumptions of policymakers, building or participating in international frameworks, and consistently resisting framings that treat coordination as naive or premature. These are real contributions to conditions that haven’t yet been created — and creating them is the work of the decade, not the quarter.
On timing
Leverage points don’t stay equally accessible over time. The legal loop is most accessible while courts are still setting precedent — before one appellate standard gets entrenched. The information loop is most accessible before opacity gets institutionalized in vendor contracts and regulatory carve-outs. The coordination window is most accessible before capability is far enough ahead of oversight that major actors have no face-saving path back. These windows are not closed. Some are narrowing. The tier cards that follow note where timing is particularly acute.