Can governments effectively regulate artificial intelligence?

The Alberta government is evaluating how to regulate AI to address harms from deepfakes and misinformation. CBC interviewed Sam Jenkins, founder of Punchcard Systems, who discussed the risks and opportunities of AI and the policy challenges around content integrity and misinformation.

Analysis

Regulatory attempts to curb deepfakes and misinformation create a predictable demand shift: buyers of detection, provenance, and identity tooling at scale. Expect cloud consumption and GPU cycles for watermarking/detection models to ratchet up, favoring hyperscalers and silicon vendors; conversely, ad-driven platforms will see margin pressure from higher moderation and compliance spend (we model a 2-5% incremental opex hit for top-5 social platforms over 12-24 months). Smaller consumer-facing startups without embedded compliance stacks face either accelerated M&A or capitalization stress as compliance becomes a fixed cost.

Timing is staggered: immediate media and political attention will produce headlines and pilots in days-weeks, legislation and enforceable rules will take months, and material capex/opex reallocation (cloud/GPU spend, hiring of trust & safety teams) will be measurable over 12-36 months. Catalysts to watch are (1) any high-profile election-cycle deepfake with viral reach, (2) provincial bills becoming templates for federal rules, and (3) large-platform policy changes or class-action litigation — each can compress the adoption curve by 3-6 months.

A contrarian angle: the market currently prizes pure-play detection startups as the “solution,” but enforcement fragmentation (province-by-province rules, cross-border data constraints) increases compliance complexity and raises the value of incumbents that can bundle identity, cloud, and moderation services. That structure favors integrated players (cloud + security + identity) and creates a near-term arbitrage: buy infrastructure/security exposure and avoid single-product detection names that will be acquisition targets or squeezed on pricing.

Key Decisions for Investors

• Long NVDA (NVIDIA) via 9–15 month call spread or outright stock exposure — thesis: incremental GPU demand from watermarking/detection increases data-center utilization. Target +30–60% in 12 months; downside = premium paid/market drawdown. Size as a thematic overweight (5–7% of tech exposure).
• Pair trade: Long CRWD (CrowdStrike) vs Short META (Meta Platforms) over 6–12 months — cybersecurity vendors capture recurring ARR uplift from provenance/monitoring; large social platforms absorb higher moderation costs and regulatory uncertainty. Target spread improvement 15–25%; stop-loss if CRWD underperforms Nasdaq by >12% or META rises >20% on ad-recovery news.
• Overweight MSFT (or AMZN/AWS) for 12–24 months — hyperscalers gain attach rates from compliance tooling, cloud ingestion, and enterprise contracts. Expect lower-beta upside (~15–25%) with defense against idiosyncratic regulatory shocks; maintain hedges for macro risk.
• Tactically short or buy puts on small-cap pure-play detection firms (e.g., VERI-sized market caps) for 6–12 months — these names are acquisition targets or face pricing pressure as incumbents bundle solutions. Target 30–50% downside; keep position sizes small and time-limited due to illiquidity.