Microsoft and OpenAI broke – now they are ready to fight

At Microsoft’s annual Build conference on Tuesday, the company announced new or expanded AI solutions, including advanced software, internal models, a cybersecurity tool, and OpenClaw-esque AI agents. All of this news added up to a clear message: Microsoft is positioned to be one of the biggest players in AI, and it’s finally doing just that.

For years, Microsoft’s AI business has leaned heavily on its first and only partnership with OpenAI. But a family full of drama gradually became a situation, and the two well separated at the end of April (although Microsoft is still the main partner of OpenAI – for now). This year’s build had the vibe of a newly divorced person posting a thirst trap on Instagram. “It’s always exciting to be at a developer conference during a time of great change,” Microsoft CEO Satya Nadella said on Tuesday, adding that events like this are about “receiving new opportunities.”

AI CEO Mustafa Suleyman, in an interview with Seasideto put it bluntly.

“The goal is to prove that we can be one of the top four labs in the world,” Suleyman said. “There are three labs that are important, Google DeepMind, OpenAI, and Anthropic. We’re not one of them at the moment, and that’s always my goal. That’s why I came here. I want to build the best examples of frontiers in the world, in great numbers, and to do that, we have to prove that we can do everything we need from the ground up, and not just go to others.”

One of Microsoft’s first moves at Build was to play around with AI models. Suleyman unveiled the MAI-Thinking-1, the company’s first model, along with six other new models that focus on image, voice, text, and text. Microsoft he said the intermediate version of MAI-Thinking-1, which will be marketed primarily to business customers, is “built from the ground up to use math, code, and deploy real businesses around the world.” Microsoft is years behind OpenAI and Anthropic here; OpenAI began to produce conceptual models at the end of 2024. But Suleyman emphasized its performance on indicators such as documentation and its cost, saying that it was cheaper than similar OpenAI for other applications – a big thing in the years of AI money squeezewhich has inspired many complaints by customers.

Although Microsoft has had years to learn more from OpenAI, Suleyman was sure to say that its development did not include melting, meaning that it was not trained to use a particular type of AI for the company. If MAI-Thinking-1 is good, Microsoft doesn’t want people to think it’s because of OpenAI.

Suleyman said Seaside that for Microsoft, “the most important moment was the renegotiation of our partnership with OpenAI. This means that we are allowed to train models on a large scale and pursue advanced intelligence entirely with our IP, with our data, unpolished, trained from scratch.”

Nadella also highlighted Microsoft’s recently launched cybersecurity tool MDASH, which he said brings together 100 AIs to find exploitable bugs “more than any other brand.” It was good to dig into the Claude Mythos Preview, which Anthropic launched in April A lot of fear and excitement – and opportunity to add to before I built it. OpenAI has its approach that focuses on cybersecurity also, and all three companies can use their contributions to work in the government and business markets that are needed in the court.

Microsoft is in deep trouble with AI assistants. The a popular open source platform OpenClaw showed a potential way forward for AI agents, and beyond OpenAI hired quickly Its creator, Peter Steinberger, Microsoft (among other companies) is trying to find. One of its main goals is to make OpenClaw work better with Windows. At Build, Nadella said he was very committed to supporting OpenClaw, and Microsoft employees talked to developers in the audience about how they are using it.

Steinberger himself made a surprise appearance to the large audience, taking the stage to praise how OpenClaw has strengthened its security and made users trust them. “All I could hear was, ‘Peter, I love my Claw, but can I use it at work?'” Steinberger said. “You can run OpenClaw inside your company now, and we also created a plug-in.” Steinberger said that if one trusts Copilot, Codex, or another company’s platform, users can run OpenClaw on top of them through Windows.

But Microsoft is also promoting a “premium” Copilot program that includes OpenClaw-esque agents. The top program is the focus of OpenAI right now – President Greg Brockman is leading development that integrates ChatGPT, the Codex coding platform, and the Atlas browser. Microsoft’s approach is similar, bringing together the various existing Copilot AI assistants. Its assistants, called “Autopilots,” are designed to act as a user interface. Cassidy Williams, GitHub’s senior director of public relations, called Copilot “your desktop development and service center,” showing how multiple collaborators can work together as a developer. (Increasingly, Williams demonstrated how to accept or reject code changes by flashing his thumb or thumb on the computer’s camera.)

Pilots are designed specifically to appeal to business customers — Nadella called them “independent, long-term workers who target all businesses.” The first one Microsoft will offer is “Scout,” written as “your personal assistant,” but customers can create and customize their own. Autopilot contributors can check email, join group chats in Groups, view the calendar, and send daily updates, among other things. Therefore, the staff on the Build stage repeatedly emphasized the equipment and safety of the Copilot – apparently to reassure customers who may have heard. terrible news of tools such as OpenClaw.

Suleyman made sure to emphasize, repeatedly, Microsoft’s “human intelligence” as “AI that puts humanity first” – part of the latest AI industry. Changes in the value of AGI to make it sound less dangerous in the time people are in pushing back against the AI more than ever.

Nvidia CEO Jensen Huang, another speaker who is known for working with OpenAI, appeared via video to find out how Nvidia’s RTX Spark chip supports Microsoft’s AI goals. “I can be walking and I’m on the phone and I can send text messages on my PC … and it can turn on devices on the PC,” Huang said. “The idea that the PC evolved from a computer to a human AI is very interesting.”

Microsoft spent years betting on OpenAI, and in some ways, it has lagged behind in the AI ​​race. But as OpenAI and other competitors turn to business for funding, it has some obvious advantages. Microsoft already has many customers and, compared to other AI companies, a reputation for safety and security. And like Googleit also has deep pockets, extensive computer hardware, and a variety of funding options, meaning it can bet big without risk.

Suleyman said Seaside“There are a lot of people who are kind of chasing the initial valuation or the IPO, so we can work with a little bit of humility and long-term optimization.” He added: “We have the money to buy Anthropic (samples) when we need to. We have access to Azure with 11,000 models, so people can use whatever they want whenever they want, but that buys us time to do it from scratch.”

At the same time, there are many unanswered questions here. Microsoft has announced great success with the development of its seven new models, but it does not always translate to the real world, and the new model. which drags on for a week or two can fall quickly. Advanced AI software is an untested concept. And Microsoft’s entry into the crowded but complex AI market is something we’ve never seen before. There are many places for his promises to fall.