OpenAI, Microsoft rift hinges on how smart AI can get

The contract between the tech partners, who have been locked in acrimonious negotiations, stipulates that when OpenAI’s systems reach “artificial general intelligence," or AGI, the startup will be able to limit Microsoft’s access to its future technology. Microsoft is fighting hard to prevent that.

Many AI experts see AGI as the point at which generative AI systems achieve humanlike intelligence, but OpenAI and Microsoft are at odds over the issue. OpenAI executives including Sam Altman believe they are close to being able to declare that their AI tools have achieved the AGI level of proficiency, according to people familiar with the matter.

Microsoft Chief Executive Satya Nadella has expressed skepticism that reaching such a benchmark is possible. Their disagreement mirrors a debate among Silicon Valley’s elite about just how sophisticated cutting-edge tools can become.

“We have a long-term, productive partnership that has delivered amazing AI tools for everyone. Talks are ongoing, and we are optimistic we will continue to build together for years to come," representatives for both companies said in a statement.

The two companies are working to renegotiate their commercial agreement as OpenAI seeks to convert itself into a for-profit company. The conversion will unlock tens of billions of dollars in new funding to advance its artificial intelligence tools.

Today, Microsoft requires OpenAI to sell its AI software exclusively through Azure—a restriction from doing business with other cloud providers that few other AI developers face. Microsoft, which integrates OpenAI’s latest technology across many of its major products, also controls OpenAI’s intellectual property.

In recent negotiations, Microsoft has indicated that it is willing to accept an equity stake of about 35% in the new for-profit company, people familiar with the matter said.

The AGI divide looms over the discussions between the two companies.

Altman has said AGI, which his company defines as “highly autonomous systems that outperform humans at most economically valuable work," is just around the corner.

Microsoft’s Nadella has been more critical. “Us self-claiming some AGI milestone, that’s just nonsensical benchmark hacking," he said on a popular tech podcast in February. The comments surprised some OpenAI officials, who said Nadella was once described by Altman as an “AGI believer."

OpenAI executives have discussed the possibility of declaring AGI through an AI coding agent that exceeds the capabilities of an advanced human programmer, people familiar with the matter said.

The contract only requires that OpenAI’s board declare AGI in good faith, though Microsoft could easily sue the company in response, risking a drawn-out legal battle.

OpenAI can also declare a higher tier of AGI, called “sufficient AGI," when its AI systems are financially capable of paying Microsoft the future profits to which it is entitled. Microsoft has to sign off on sufficient AGI, at which point OpenAI would get the right to sell or license all of its technology to other cloud providers.

Microsoft is restricted from developing AGI on its own under the companies’ contract, which runs through 2030, according to people familiar with the matter. It is OpenAI’s largest outside shareholder and has the ability to veto any major funding decision that dilutes its equity.

Some Microsoft executives objected to including the AGI clause when the partnership was first negotiated in 2019, believing that it was arbitrary and unenforceable. But the company was so behind on AI at the time that Nadella agreed to it, according to people familiar with the matter.

Microsoft is hoping to remove the clause entirely from the contract as part of its recent negotiations, or secure exclusive access to OpenAI’s IP even after AGI is declared.

Write to Berber Jin at berber.jin@wsj.com