OpenAI just published its roadmap for how governments should regulate the thing it's building, which is either public service or the fox designing the henhouse.

The Summary

  • OpenAI released policy recommendations for how society should adapt to AI-driven change, presented by Chief Global Affairs Officer Chris Lehane
  • The company is positioning itself as the architect of AI governance, not just AI technology
  • This marks a shift from "move fast and break things" to "move fast and write the rulebook"

The Signal

OpenAI's Chris Lehane went on Bloomberg Tech to present the company's vision for AI policy, framed around "ensuring AI benefits everyone." The timing matters. We're past the experimental phase where AI was a research curiosity. We're now in the phase where OpenAI is publishing policy recommendations about the societal changes its own products are causing.

This is a familiar playbook. Hire a chief global affairs officer. Release thoughtful-sounding policy papers. Get ahead of regulation by suggesting what regulation should look like. The subtext is always: "Please regulate us in ways that don't slow us down or help our competitors catch up."

The real question is whether OpenAI's recommendations address the immediate displacement happening right now. Customer service reps, paralegals, junior analysts, content moderators. These jobs aren't disappearing in some distant future requiring new policy frameworks. They're disappearing this quarter. Policy discussions about "ensuring AI benefits everyone" tend to skip over the part where "everyone" includes people whose work an LLM can do for $0.02 per thousand tokens.

The Implication

Watch what OpenAI recommends versus what it lobbies for. Policy recommendations are public relations. Lobbying is strategy. If you're in a knowledge work job that involves processing information and generating text, don't wait for policy to protect you. The companies building these tools are writing the rules, and the rules will optimize for building faster, not protecting slower. Learn to use agents, or learn to do something agents can't.

Source: Bloomberg Tech