Artificial Intelligence: Free Markets, Bad Actors, and Economic Transformation

Summary

• Generative AI was in early development within frontier labs until approximately four years ago.
• AI models were released to individuals and companies starting around 2022.
• Competition among AI labs drove exponential improvements in model capabilities.
• By early 2026, ChatGPT reported over 900 million weekly users, with roughly 92% of the Fortune 500 utilizing it.
• Workplace AI seats reached over seven million by mid-2026, representing a ninefold increase from the prior year.
• Frontier models evolved from handling minute-long tasks to managing tasks lasting roughly two hours for GPT-5.
• Some capacity measures for AI models are improving at a rate close to tenfold annually.
• Risks associated with AI include fraud, cyberattacks, deepfakes, strain on local power & water from data centers, & job displacement.
• No federal rules currently exist for AI developers, though California requires large model developers to disclose safety practices.
• Twenty-six state attorneys general have requested congressional action regarding AI regulation.
• President Trump signed a voluntary accord with AI leaders asking the industry to police itself through internal controls, outside audits, & an oversight board.
• The accord is described as morally binding rather than legally codified at this time.
• Studies forecasting increased productivity & economic growth typically assume little or no government regulation.
• Polls indicate majorities in both political parties favor increased legislative focus on AI.
• AI-related topics are expected to influence midterm elections & the 2028 election cycle.
• Some consultants project potential productivity gains of several percentage points from AI adoption.
• Federal researchers frame AI as raising the economy's speed limit through innovation in how work is performed.
• Optimism regarding AI is tempered by worries about bad actors & potential volatility in AI-related company valuations.
• The base case suggests the AI buildout will not slow down despite regulatory uncertainty.