Ex-OpenAI CTO Mira Murati launches AI startup, recruits top talent from rivals

Former OpenAI CTO Mira Murati has launched an AI startup, Thinking Machines Lab, escalating the fight for top talent and reshaping competition in the fast-evolving sector.

About 30 researchers and engineers at Thinking Machines Lab previously worked at competitors including OpenAI, Meta, and Mistral, Reuters reported. Former OpenAI employees alone make up about two-thirds of the company.

On its website, Thinking Machines Lab said AI advancements have outpaced scientific understanding, with knowledge concentrated in top research labs, restricting public discourse and accessibility.

“Despite their potential, these systems remain difficult for people to customize to their specific needs and values,” the company added. “To bridge the gaps, we’re building Thinking Machines Lab to make AI systems more widely understood, customizable, and generally capable.”

Former OpenAI employees who have joined Thinking Machines Lab include Barret Zoph, an AI researcher who left OpenAI with Murati in late September, and John Schulman, an OpenAI co-founder who joined Anthropic in August to focus on AI safety.

More OpenAI employees are expected to join Thinking Machines Lab, according to the Reuters report.

AI talent war intensifies

The shift underscores growing competition in the generative AI sector, with top talent moving between established firms and emerging startups.

Analysts expect this trend to continue, driven by demand for cost-effective models and innovative AI solutions.

“Executives and senior technology leaders from major pioneering companies setting up their own ventures is not uncommon in the industry,” said Sharath Srinivasamurthy, associate vice president of Research at IDC. “Also, in the highly dynamic world of AI, talent retention is challenging, and companies should have contingency plans in place to address these kinds of developments.”

In the short term, OpenAI may face challenges in retaining mindshare and employee confidence. However, ongoing investments, backing from Microsoft, and initiatives like Project Stargate reinforce its position as one of the most desirable companies in the AI sector.

“Those who were not aligned with Sam Altman’s (and investors’) vision for OpenAI have left, making the company free of internal politics, power struggles, and other distractions that could slow its progress,” said Neil Shah, partner and co-founder at Counterpoint Research. “In that sense, last year’s exodus may have been a blessing in disguise for OpenAI.”

Hurdles ahead

Developing advanced AI models demands substantial infrastructure and a highly skilled research team, making it a long-term effort.

In contrast to companies like DeepSeek, which aim to challenge OpenAI and other generative AI leaders within a few years, Thinking Machines Lab appears to be taking a longer-term approach.

“While current systems excel at programming and mathematics, we’re building AI that can adapt to the full spectrum of human expertise and enable a broader spectrum of applications,” Thinking Machines Lab noted on its website.

However, it may face significant challenges in the crowded AI startup market, especially given the complexity and long-term nature of AI alignment research, said Charlie Dai, VP and principal analyst at Forrester.

“By leveraging its niche focus, emphasizing ethical AI, and building strategic partnerships, the lab can carve out a unique position and capitalize on the growing importance of AI alignment,” Dai said. “Success will depend on its ability to balance long-term research goals with practical, commercially viable applications.”

However, perhaps the biggest challenge Thinking Machines Lab faces is the apparent ceiling on the utility large foundation models can offer the average user. “The goal of AI alignment with human values is good in theory, but difficult to quantify or qualify as a discrete advantage,” said Hyoun Park, CEO and chief analyst at Amalgam Insights. “In theory, the ability to better align AI to human values could theoretically create a better UX that is more aligned to both help people solve problems or to solve problems in a more human fashion.”