Risk 8: AI eats its users

Under pressure to generate revenue and unconstrained by guardrails, a number of leading AI companies will adopt business models in 2026 that threaten social and political stability—following social media's destructive playbook, only faster and at greater scale.

We remain bullish on AI's revolutionary potential. Today's frontier models reason through complex problems, show their work, and are embedded in coding, research, and knowledge workflows. The hyperscalers are offloading large chunks of software development to AI, accelerating their own R&D cycles. In biotech and materials science, AI is opening new research pathways—though commercial breakthroughs remain mostly ahead of us. Hundreds of millions of people now use chatbots daily for everything from drafting emails to debugging code and learning new skills. This is real, and it's just the beginning.

But AI can't live up to investors' expectations in the short term. Even after hundreds of billions of dollars of investment, the most advanced models still hallucinate. Their capabilities are jagged: dazzling at some tasks, unreliable at others (and often unpredictably so). That inconsistency makes them hard to deploy in high-stakes applications where errors are costly. Business adoption has been uneven, with only about 10% of US firms using AI to produce goods and services, according to the Census Bureau. Many companies report significant productivity gains, but surveys suggest most have yet to see meaningful bottom-line impact. Real productivity increases will arrive through wide diffusion of the technology across the economy, but that takes time. Yet markets have priced in revolution, not evolution.


To read the full risk, download the report.