Six years ago, OpenAI promised to give AGI away if someone else built it first — now they've quietly deleted that promise from their principles.
The Summary
- OpenAI published updated principles that dramatically shift from their 2018 charter, notably reducing AGI mentions from 12 to 2 and reversing their stance on competition
- The original nonprofit promised to assist competitors if they were close to AGI; the new for-profit version explicitly embraces competitive advantage
- This isn't a policy tweak — it's documentation that the company Elon Musk co-founded to democratize AI has fully morphed into a standard Silicon Valley growth machine
The Signal
OpenAI's 2018 charter included a commitment most people forgot: if another organization got close to building AGI before them, OpenAI would stop competing and start assisting. That sentence is gone. So is most of the AGI talk entirely.
The company now mentions artificial general intelligence twice in their principles document, down from 12 times in 2018. This isn't about brevity. It's about business reality catching up to founding mythology.
"We believe society needs to contend with each successive level of AI capability."
The new framing is cleaner, more defensible, and conveniently aligns with shipping products on quarterly timelines. AGI is a useful north star for recruiting researchers and raising capital, but it's a terrible framework for building a company that needs revenue this year. So they've pivoted to "iterative deployment" of increasingly capable AI, which is just product management dressed up as philosophy.
- 2018: "We'll help competitors if they're close to AGI"
- 2026: Full competitive mode, no collaboration promises
- Translation: We're a normal company now
The competition reversal is even starker. The original charter explicitly said OpenAI would assist rival labs nearing AGI breakthroughs. The new principles? Silent on collaboration, loud on maintaining "cutting edge" capabilities. This matters because former OpenAI researchers who left over these exact tensions — including co-founder Ilya Sutskever — are now building competing labs like Anthropic with the old OpenAI ethos.
The Implication
Watch what happens when mission-driven nonprofits take billions in venture capital and convert to for-profit structures. The principles don't update overnight, but they do update. OpenAI's evolution from "we'll give this away" to "we'll maintain competitive advantage" took six years and one awkward Sunday blog post.
If you're building in the agent economy, this is your reminder that partnership promises and open research commitments have half-lives. The companies racing toward Web4 will optimize for survival, not symmetry. Plan accordingly.