← Back to feed

Sam Altman published a five-point principles statement for OpenAI on April 26. It opens with a framing that his critics have used against him for years: “Power in the future can either be held by a small handful of companies using and controlling superintelligence, or it can be held in a decentralized way by people. We believe the latter is much better.”

OpenAI’s current valuation: $852 billion. Its most recent funding: $122 billion, the largest private capital raise in history. Its market position: the dominant AI consumer product (ChatGPT), the leading agentic coding platform (Codex), and an agreement to supply AI to US classified national security environments.

The five principles:

1. Democratization. “We will resist the potential of this technology to consolidate power in the hands of the few.” Key decisions about AI must go through democratic processes, not be made by AI labs alone.

2. Empowerment. AI should allow people “to achieve their goals, learn more, be happier and more fulfilled.” Users get broad autonomy; harms are minimized, not eliminated.

3. Universal Prosperity. Governments may need “new economic models to ensure everyone can participate in the value creation in front of us.” OpenAI’s aggressive capex (“buying huge amounts of compute while our revenue is relatively small”) is explicitly framed as a bet on cheap universal access.

4. Resilience. No lab can ensure a good future alone. Iterative deployment is the strategy; collaboration with governments and international agencies on alignment and societal problems is the mechanism.

5. Adaptability. OpenAI will update its positions as evidence changes and will be transparent when its operating principles shift.

The Document’s Subtext

The statement is not new policy. It’s a public articulation of principles that have internally guided OpenAI — made external at a specific moment: after the $122B round, after the GPT-5.5 launch, after Codex became a $100/month desktop agent, after OpenAI signed a classified Pentagon deployment agreement and drew scrutiny over whether its guardrails would hold in national security contexts.

The Altman document deploys the same critique as his critics — power concentration is bad — and positions OpenAI as the cure rather than the cause. The mechanism for achieving democratization is, per the document, building more infrastructure, more compute, lower costs. The result: Altman explicitly frames scale as the route to decentralization.

That’s an argument that can be made. Prices for frontier AI access have fallen 97% in three years. GPT-4-class capability is now available for under $1/M tokens. The argument that building massive supply drives access down is empirically supported at the input cost level.

Where it strains: governance. Universal Prosperity principle states “key decisions about AI” must involve democratic processes. The most consequential AI decisions — what models get released, at what safety thresholds, to whom, with what restrictions — are currently made by OpenAI’s board and leadership without any formal democratic process.

What the Document Does Not Address

The statement is written entirely in the future tense about desired outcomes. It does not:

• Specify any governance mechanism that limits OpenAI’s own power
• Define what would constitute a failure of democratization
• Commit to any hard safety floor beyond “we will not get everything right, but we will learn quickly and course-correct”
• Explain what happens if iterative deployment produces a harm that can’t be walked back

The Preparedness Framework, published separately in April 2025, covers catastrophic risk categories (biological, cyber, self-improvement). The principles document published April 26 is a values statement, not a governance document.

Timing

Altman signals the document is written with the near future in mind: “We are heading into a very impactful phase as the technology continues to improve.” GPT-5.5 was released three days earlier. Mythos is in preview. The company has publicly stated it expects to reach AGI within the current strategic horizon.

Publishing a principles document now reads as preparation — staking a public position on what OpenAI stands for before the next capability jump makes that positioning harder to establish credibly. Whether the five principles survive contact with that next jump is the question the document conspicuously does not answer.