Microsoft and OpenAI clash over future access to advanced AI

Microsoft and OpenAI clash over future access to advanced AI

The high-profile collaboration between Microsoft and OpenAI—once billed as an AI collaboration template—is now being put through a test of its mettle in one treacherous, disputed term: artificial general intelligence, or AGI, the Wall Street Journal reported.

While the firms negotiate the terms of their partnership, AGI has become the focal fault line. OpenAI has the ability under their current arrangement to restrict Microsoft's access to subsequent AI models upon announcement of AGI—something OpenAI thinks may become real shortly. Microsoft is resisting with all its might, challenging the benchmark itself as well as attempting to maintain access to the revolutionary technology it subsidized.

OpenAI thinks AGI is within reach, Microsoft resists

OpenAI defines AGI as "highly autonomous systems that outperform humans at most economically valuable work." Estimates suggest CEO Sam Altman believes the company's models are approaching this level of capability, perhaps through a coded agent that can beat even very capable human programmers.

Microsoft CEO Satya Nadella has remained sceptical too. "Us self-claiming some AGI milestone, that's just nonsensical benchmark hacking," Nadella said this year. OpenAI was taken aback, as Nadella had been referred to as an "AGI believer" by Altman in the past.

The importance of the AGI announcement is more than semantic. Once OpenAI has announced AGI, they can restrict Microsoft access to models based on their 2019 deal. That has prompted Microsoft to request modifications to their deal, such as removing the AGI section or providing continuous access even after AGI has been reached.

Power struggle for IP, revenues, and exclusivity

The AGI controversy is but one aspect of the challenge to the partnership. OpenAI is reported to be transforming itself into a for-profit enterprise, putting billions of dollars of new capital on the table. Microsoft reportedly is willing to provide up to a 35% interest in the new company but would prefer to retain its current benefits—single-source access to OpenAI's models through Azure and intellectual property control.

OpenAI is otherwise prohibited from reselling its models across other cloud providers. Microsoft wants to maintain that exclusivity even as OpenAI negotiates more commercial freedom.

In the meantime, OpenAI also has the authority to issue "sufficient AGI"—a higher stage where its models would be able to produce sufficient profitability to reward Microsoft its royalties. That announcement would authorize OpenAI to license its tools on other platforms but subject to Microsoft's approval.

There are legal risks if AGI is classified

In the present deal, it takes just a good-faith declaration by the board of directors at OpenAI to trigger the AGI clause. But Microsoft can sue if it believes the decision was made in bad faith, and that would involve a high-profile lawsuit. Some of Microsoft's executives opposed the AGI clause when it was first added in 2019 but acceded to it to acquire a stake in AI.

Microsoft, which is now one of the globe's most powerful AI players after its OpenAI takeover, is not going to bow anytime soon. With their deal set to expire in 2030, both firms now grapple with a tricky mix of business ambition, technical competition, and intellectual disagreement on what achieving human-level AI actually entails.