AI products are defective and dangerous. Why are we using them at all?

The article argues that AI products are being adopted without the safety and accountability standards applied to physical goods, highlighting widespread hallucinations, misleading outputs, and legal risk. It cites examples including Air Canada's chatbot liability case, Microsoft’s Copilot warnings, and alleged AI-generated work in a $1 million Deloitte Canada report, underscoring growing regulatory and litigation exposure. The piece calls for stronger enforcement under product liability, privacy, and deceptive marketing rules.

Analysis

The key market implication is not that AI quality is bad — that is already priced in — but that liability is migrating from model vendors to enterprise adopters faster than governance frameworks are catching up. That shifts the economic burden from capex to opex: every incremental AI deployment likely creates a hidden stack of verification, legal review, and customer remediation costs, compressing realized ROI for software-heavy buyers over the next 6-18 months.

This is most negative for companies selling AI as a productivity layer inside core workflows, where error tolerance is near zero. Microsoft and Apple face different exposures: Microsoft is more vulnerable to enterprise pullback or slower monetization of Copilot-style upsells, while Apple’s risk is reputational and regulatory if AI is embedded into consumer-facing interfaces without clear performance disclosure. Target-like retailers are an underappreciated second-order loser because AI agents can create pricing, fulfillment, and customer-service disputes that are operationally expensive even if direct legal damages are small.

The contrarian point is that regulation may ultimately be bullish for the largest incumbents. If compliance standards harden, smaller AI-native competitors with weak legal budgets and limited indemnity capacity will be squeezed first, while MSFT and AAPL can absorb the overhead and market themselves as “safer” defaults. So the near-term trade is not to short AI broadly, but to short the margin expansion narrative in names where AI monetization is still aspirational and liability is underestimated.

Key Decisions for Investors

• Short MSFT on a 1-3 month horizon via put spreads; thesis is slower Copilot adoption and higher support/compliance costs offsetting near-term AI monetization expectations. Risk/reward favors defined-risk downside if the market reprices enterprise willingness to pay for AI add-ons.
• Short TGT or buy put spreads into the next consumer/retail print; AI-driven customer-service and shopping errors raise friction costs and dispute resolution expense. Best entry is on strength after any AI-commerce optimism, with limited upside to the stock but meaningful downside if management mentions higher service costs.
• Long AAPL vs short a basket of AI-native software vendors with weak balance sheets over 6-12 months; if disclosure and reliability standards tighten, large incumbents with distribution and cash flow should outperform speculative AI monetizers. This is a quality/regulatory pair trade rather than a pure AI short.
• For event-driven traders, buy small MSFT and AAPL downside protection into product-launch windows; headline risk around AI feature rollouts can create abrupt multiple compression if any user-facing failure goes viral. Use 3-6 month tenors to capture catalyst risk without paying for too much decay.
• Avoid shorting AC.TO aggressively; the legal precedent here is directionally helpful for accountability, but the direct P&L impact is limited and any weakness is likely more sentiment than fundamental. Better expressed as a relative short against names with higher AI monetization hype.