Sam Altman publishes 'Planning for AGI and beyond'

OpenAI chief executive Sam Altman published “Planning for AGI and beyond,” the company’s most detailed public statement of its mission and strategy since its founding charter. It restated OpenAI’s goal as ensuring that artificial general intelligence — systems generally smarter than humans — “benefits all of humanity,” and laid out principles for how the company intended to get there.

The essay’s central argument was for iterative deployment: releasing progressively more capable systems into the world, rather than developing AGI in secret and deploying it all at once. Altman wrote that this let society, regulators and OpenAI itself adjust gradually, reducing the risk of a “one shot to get it right” scenario, while acknowledging that the approach carries its own danger if a capability jump turns out to be too large to correct for after release.

A misaligned superintelligent AGI could cause grievous harm to the world; an autocratic regime with a decisive superintelligence lead could do that too.

— Sam Altman, “Planning for AGI and beyond”

The essay named short-term priorities including building “increasingly aligned and capable” systems, working to broadly and fairly distribute AGI’s benefits and governance, and navigating risks including misuse, accidents and societal disruption. It stopped short of specific commitments — no compute thresholds, no external triggers for pausing development — functioning instead as a statement of intent against which OpenAI’s later, more concrete safety and deployment policies would be measured and, by critics, judged.

Published three weeks before GPT-4’s release, the essay set the terms OpenAI used to justify continued scaling even as it acknowledged catastrophic risk — a framing, contested throughout the field, that continuing to build more capable systems is itself the responsible course.