A contract term that once sounded remote is now central to one of technology’s most important partnerships. OpenAI’s agreement with Microsoft includes a clause that could limit Microsoft’s access to OpenAI’s future technologies if OpenAI’s board declares that the company has developed artificial general intelligence, or AGI.
That issue has become more sensitive as the companies renegotiate their agreement while OpenAI prepares a corporate restructuring. Microsoft has invested more than $13 billion in OpenAI, and the question is no longer only technical. It is also about control, timing, leverage, and who gets access to the most advanced systems if OpenAI says a major threshold has been crossed.
The contract clause at the center of the dispute
OpenAI says in a blog post about its corporate structure that AGI “is excluded from IP licenses and other commercial terms with Microsoft.” OpenAI defines AGI as “a highly autonomous system that outperforms humans at most economically valuable work.”
According to the Financial Times, Microsoft is pushing to remove the clause and is considering walking away from the deal entirely. Microsoft wants continued access to OpenAI’s models even if OpenAI declares AGI before the partnership ends in 2030.
One person familiar with the partnership discussions told WIRED that Microsoft does not believe OpenAI will reach AGI by that deadline. Another source close to the matter described the clause as OpenAI’s ultimate leverage. Both sources were granted anonymity to discuss private negotiations.
The clause matters because an AGI declaration could change Microsoft’s rights. One source familiar with the discussions said OpenAI’s board can unilaterally decide the company has reached AGI as defined in its charter. That would immediately cut Microsoft off from accessing the technology or revenue derived from AGI, while Microsoft would still retain rights to everything before that milestone.
Why an unreleased paper became sensitive
Late last year, the AGI issue also became tied to an internal OpenAI research paper titled “Five Levels of General AI Capabilities.” The paper outlines a framework for classifying progressive stages of AI technology.
Sources familiar with the matter told WIRED that the paper’s specific assertions about future AI capabilities could have complicated OpenAI’s ability to declare AGI. That matters because such a declaration could become a point of leverage in negotiations with Microsoft.
OpenAI spokesperson Lindsay McCallum described the work differently. “We’re focused on developing empirical methods to evaluate AGI progress—work that is reproducible, measurable, and useful to the broader field,” McCallum said in a written comment to WIRED. “The ‘Five Levels’ was an early attempt at classifying stages and terminology to describe general AI capabilities. This was not a scientific research paper.” Microsoft declined to comment.
Bloomberg previously reported on the existence of the “Five Levels” and that OpenAI planned to share the scale with outside investors, though it was considered a “work in progress.” OpenAI CEO Sam Altman and chief research officer Mark Chen have discussed the five levels of AI capabilities in interviews since.
What the five levels tried to describe
A version of the paper dated September 2024, viewed by WIRED, details a five-step scale for measuring how advanced AI systems are. It cited other research claiming many OpenAI models at that point were at Level 1, defined as “An AI that can understand and use language fluently and can do a wide range of tasks for users, at least as well as a beginner could and sometimes better.”
The paper also said some models at the time were approaching Level 2. The authors defined that as “An AI that can do more advanced tasks at the request of a user, including tasks that might take an hour for a trained expert to do.”
Instead of giving one single definition of AGI, the paper avoided treating AGI as a simple yes-or-no category. It used a spectrum of capabilities to describe increasingly general and capable AI systems. That approach may be useful for discussing progress, but it can also create complications when a contract turns on whether AGI has been reached.
The paper did not predict when OpenAI systems would reach each level. It did, however, predict how each increase in capability could affect areas including education, jobs, science, and politics, while warning about new risks as AI tools become more powerful and independent.
The negotiation stakes are broader than terminology
The Wall Street Journal reported that OpenAI has considered whether to invoke the clause based on an AI coding agent. The Journal also reported that talks became so tense that OpenAI discussed whether it should publicly accuse Microsoft of anticompetitive behavior.
Another layer is the contract concept of sufficient AGI, added in 2023. According to a source familiar with the discussions, that term defines AGI as a system capable of generating a certain level of profit. If OpenAI says it has reached that benchmark, Microsoft must approve the determination.
The same source said the contract bars Microsoft from pursuing AGI on its own or through third parties using OpenAI’s IP. That makes the definition of AGI more than a research debate. It affects future access, revenue, and what each company can do next.
Altman has said he expects to see AGI during Donald Trump’s current term, according to the source familiar with the discussions. In a podcast with YCombinator president and CEO Garry Tan in November, Altman said the company’s o1 model could be defined as Level 2, and that he expects OpenAI will reach Level 3 “faster than people expect.”
Why the paper stayed unpublished
Last July, a coauthor presented the Five Levels research at an internal event where teams highlighted important projects for research-wide awareness. Multiple sources said the work appeared to be in final stages, and that OpenAI had hired a copy editor to finalize it late last year while also generating visuals for a blog announcing the paper.
Several sources told WIRED that OpenAI’s partnership with Microsoft was cited internally as one reason to delay publication. Another source said discussions with Microsoft were often “mentioned as a blocker for putting the paper out.”
McCallum disputed that framing, saying “it’s not accurate to suggest we held off from sharing these ideas to protect the Microsoft partnership.” Another source familiar with the matter said the paper was not released because it did not meet technical standards.
The result is a dispute where language, research framing, and contract rights all overlap. At a conference in early June, Altman said, “I think mostly the question of what AGI is doesn’t matter,” adding that “It is a term that people define differently; the same person often will define it differently.” In this partnership, however, the definition may matter precisely because the contract says it does.