OpenAI sets five ground rules for the AGI era

OpenAI wants the world to know it hasn't lost the plot.

CEO Sam Altman published a new principles document outlining five core commitments guiding the company's path toward artificial general intelligence: democratization, empowerment, universal prosperity, resilience, and adaptability.

Beyond high-minded rhetoric about superintelligence lifting humanity, the manifesto addresses a pointed critique: why does OpenAI keep making moves that look outright bizarre from a traditional business perspective?

Explaining the compute spending spree

Buying staggering amounts of compute while revenue remains comparatively small, aggressively pursuing vertical integration, and building datacenters worldwide might look irrational. OpenAI says these moves are entirely deliberate.

According to the manifesto, achieving shared prosperity requires driving AI infrastructure costs down so dramatically that anyone can build with frontier systems. The company argues that lower marginal costs will unlock new scientific discoveries and economic models that governments must eventually support.

On safety, OpenAI committed resources from its Foundation to tackle systemic risks, including model-assisted pathogen creation and cybersecurity threats. It reiterated its iterative deployment philosophy, openly admitting that its early panic over releasing GPT-2 weights was misplaced in retrospect, but instrumental in proving that AI labs cannot develop safety protocols in a vacuum.

OpenAI also acknowledged its own outsized footprint. As the lab grows, it pledged to remain transparent when its principles shift, noting it may occasionally have to trade off individual user empowerment for collective resilience.

Why it matters

This is less of a moral manifesto and more of a preemptive defense of OpenAI's massive capital expenditure.

For developers and users, the explicit nod to giving people broad autonomy is welcome news. But the fine print is equally clear: OpenAI is reserving the right to throttle capabilities or pause deployments whenever biosecurity or alignment risks loom large. If you are building mission-critical tools on frontier models, regulatory friction remains baked into the roadmap.

For the broader industry, Altman is framing OpenAI's infrastructure land grab not as aggressive market capture, but as a public utility effort to make intelligence cheap. That sets a daunting benchmark for rival labs, who must now justify both their commercial viability and their alignment philosophies under intense public scrutiny.

We think the document's biggest tension lies in its democratic pledge. OpenAI argues decisions shouldn't just be made by AI labs, yet it remains firmly in the driver's seat when deciding when to compromise user autonomy for safety.