I helped build Facebook and saw it go wrong. AI is headed the same way

66% of Americans support citizen panels to help set AI rules and 79% worry the government has no plan for AI-driven job loss. The author warns that tech firms racing to deploy AI ("if we don't do it, someone else will") will optimize for engagement and profit unless democratic citizen assemblies with binding authority set constraints. The piece calls for public control of AI resources and democratic governance as the lever to redirect AI wealth toward social goods (child care, retraining) and to avoid repeating social media’s harms.

Analysis

A sustained push for citizen-led governance of AI materially raises regulatory uncertainty for engagement- and ad-optimized business models while simultaneously increasing demand for auditable, compliant infrastructure. If assemblies or equivalent democratic processes produce binding constraints on profiling, targeted advertising, or automated decision-making, firms with >30-50% ad-exposure could face a 5-15% EBITDA compression over 12–24 months due to reduced personalization and higher compliance costs; conversely, certified cloud/AI-hosting providers and managed security vendors should see incremental revenue as customers migrate to vetted environments.

Timing matters: political signaling and regulatory drafts will dominate headlines in weeks–months, pilots and jurisdictional rules will appear in 6–18 months, and coordinated or binding frameworks (or formalized citizen assemblies with enforcement power) are a 1–4 year outcome. Tail risks include (a) rapid multinational coordination that forces revenue-sharing or windfall taxation on AI-derived revenues (potential >20% EPS hit for ad-heavy platforms) and (b) fragmented local regimes that benefit large incumbents able to absorb compliance costs; reversals would come from credible industry self-regulation, demonstrable safety/utility wins, or strategic national-security exceptions that soften public-control outcomes.

The consensus underestimates frictions: democratic processes rarely produce fully harmonized technical rules quickly, which paradoxically advantages vertically integrated hyperscalers and large cloud players who can internalize governance pipelines. That non-obvious second-order effect means regulation could concentrate economic value, not diffuse it — a structural factor to weight into multi-year positioning.

Key Decisions for Investors

• Long NVDA (NVDA) — buy shares or Jan-2027 calls to capture sustained AI compute demand and the premium pricing power of constrained GPU supply; time horizon 12–36 months. Risk: high near-term vol and cyclical GPU inventory; reward: 2–4x base-case on continued enterprise AI adoption, downside if datacenter capex stalls.
• Pair trade: Long MSFT (MSFT) / Short META (META) — 1:0.6 notional over 6–24 months. Rationale: MSFT benefits from enterprise/cloud migration to compliant platforms; META is most exposed to engagement/targeting headwinds. Risk/reward: asymmetric — limited upside if macro tech multiple compresses, but a 20–30% relative outperformance if regulation curtails ad personalization.
• Long cybersecurity and compliance software (CRWD, PANW) — buy CRWD or PANW shares or 12-month call spreads. Expect 15–50% upside as firms increase spend on auditability and secure hosting; downside if fiscal austerity curbs enterprise software budgets.
• Tactical hedge: Buy META 9–12 month put spread (e.g., buy 1 put / sell lower-strike put) to cap premium while protecting against a 20%+ downside from regulatory shock. Cost-limited protection aligns with a portfolio overweight to AI infrastructure names.