Source: the clause that rescinds Microsoft's access to OpenAI's most powerful tech if OpenAI develops AGI remains part of their new deal, but has been modified
The start-up reached a tentative deal with Microsoft, its biggest investor, and said it would give a $100 billion stake to the nonprofit that manages it.
Context & Ripple Effects
The AGI-triggered access provision has been the central fault line in the Microsoft-OpenAI relationship: OpenAI had explored removing it, while reports later described OpenAI as controlling the determination of AGI and “sufficient AGI.” Earlier talks over dropping the access cutoff therefore framed the clause as a bargaining lever, not boilerplate.
The tentative agreement preserves that lever in modified form while pairing it with a proposed $100 billion stake for OpenAI’s nonprofit manager. That makes the definition and consequences of an AGI threshold consequential to both the companies’ commercial relationship and OpenAI’s governance structure.
First-order effects
- Microsoft and OpenAI retain a contractual boundary under which OpenAI’s most powerful technology could cease to be available to Microsoft if OpenAI develops AGI; the modification leaves the precise scope unresolved in this report.
- OpenAI keeps an AGI-linked mechanism in the tentative deal rather than fully conceding the removal Microsoft had reportedly sought. Microsoft’s push to eliminate the clause had put that mechanism directly at issue.
Second-order effects
- The parties will need to translate the modified trigger into operational terms—especially around access, model capability, and governance—because prior reporting placed AGI and “sufficient AGI” determinations with OpenAI. The reported allocation of AGI decision-making to OpenAI gives those definitions commercial weight.
- A continuing conditional-access clause limits how completely Microsoft can treat OpenAI’s frontier output as a permanently available input, while giving OpenAI more leverage in future investment and product-access negotiations.
Third-order effects
- If frontier-model partnerships increasingly attach access rights to capability thresholds, contracts—not just technical performance—will determine who can commercialize leading models and on what terms.
- The case points to a durable tension in frontier AI: capital partners seek dependable access, while labs preserve governance-controlled escape hatches for increasingly capable systems.
The trend: Frontier AI partnerships are evolving from broad exclusive access arrangements toward negotiated, capability-contingent control over model rights.