These 9 quotes span 2022 to 2026. Murati's arc is institutional, not intellectual: she left OpenAI for her own lab, but her core view that safer AI and more capable AI advance together has held across the move and through Thinking Machines Lab's launch.
1. Why Does Mira Murati Say AI Safety and Capability Go Hand in Hand?
"For capabilities and safety, they're actually not separate domains. They go hand-in-hand."Mira Murati, Dartmouth Engineering, June 2024
Smarter AI is easier to align, not harder, Murati argues, because a more capable model can understand a negative directive and follow it. Dartmouth's AI Everywhere event gave her the platform, on the same trip when the university awarded her an honorary doctorate. The position separates Murati from researchers like Yoshua Bengio, who argue today's training methods produce models that scale unpredictably.
2. Why Did Mira Murati Leave OpenAI in 2024?
"There's never an ideal time to step away from a place one cherishes, yet this moment feels right."Mira Murati, SiliconAngle, September 2024
After six and a half years leading OpenAI's research and product organization through ChatGPT, GPT-4, and Sora, Murati stepped down in September 2024. She posted the resignation note on X the day OpenAI confirmed it was restructuring into a for-profit entity, alongside two senior research executives who left within days. Five months later, in February 2025, she announced Thinking Machines Lab.
3. What Did Mira Murati Warn About a "Race to the Bottom on Safety"?
"I think the downside is a race to the bottom on safety. That is the downside for sure."Mira Murati, Fortune, October 2023
Competition between AI labs creates pressure to cut safety corners, the central industry risk Murati has named since 2023. She gave the warning to Fortune in October 2023 as OpenAI prepared its first DevDay and developers debated how fast frontier labs were releasing models. Dario Amodei's Anthropic was founded on a similar premise: stay capable enough that safety-focused labs aren't outrun by less cautious ones.
4. How Far Did Mira Murati Say AGI Was From Being Safe in 2023?
"We're far from the point of having a safe, reliable, aligned AGI system."Mira Murati, Associated Press via SecurityWeek, April 2023
No AI system, including OpenAI's, was yet safe enough to be considered aligned with human intentions, Murati told the Associated Press in April 2023. The interview came four months after ChatGPT's debut and one month after GPT-4, when expectations of AI capability were climbing fastest. The assessment placed Murati alongside Demis Hassabis, who has argued AGI will arrive in stages rather than as a single breakthrough.
5. Does Mira Murati Think AI Should Be Regulated?
"It's not too early. It's very important for everyone to start getting involved, given the impact these technologies are going to have."Mira Murati, TIME, February 2023
Weeks after ChatGPT's launch, before most lawmakers had heard of OpenAI, Murati called publicly for AI regulation in a February 2023 TIME interview. The profile was the first to frame Murati as an engineer arguing for outside oversight rather than self-regulation. The early position prefigured the US Senate Judiciary hearings later that summer, where other AI leaders made parallel cases, and the EU AI Act finalised the following year.
6. What Mission Did Mira Murati Set for Thinking Machines Lab?
"Thinking Machines Lab exists to empower humanity through advancing collaborative general intelligence."Mira Murati, Maginative, July 2025
Where OpenAI talked about beneficial AGI, Murati's new company describes itself as building AI that works alongside people, not in place of them. She posted the mission statement on X on July 15, 2025, the day Thinking Machines announced a $2 billion seed round at a $12 billion valuation, the largest seed in startup history. Two things set the lab apart from OpenAI and Anthropic: multimodal systems, and an open-science stance Yann LeCun has championed at Meta.
7. How Does Mira Murati Compare AI to Electricity?
"We have to think about AI like electricity."Mira Murati, Greylock Intelligent Future summit, September 2022
AI needs governance like societies regulate electricity, Murati argues, not the testing-and-release cycle of a discrete product. She offered the analogy at Greylock's September 2022 Intelligent Future summit, in a discussion with Stanford's Fei-Fei Li moderated by Reid Hoffman, two months before ChatGPT launched. The view reflects Murati's belief that no single company or country should define AI alone.
8. What Did Mira Murati Tell Kara Swisher About AI's Promise?
"These technologies are amazing, and they have incredible promise. We can get this right."Mira Murati, Johns Hopkins Bloomberg Center, June 2024
When journalist Kara Swisher questioned whether OpenAI took AI's downsides seriously, Murati called the framing "a hopeless approach." The exchange happened on June 10, 2024, during the inaugural taping of Swisher's On with Kara Swisher podcast at the Johns Hopkins Bloomberg Center in Washington, D.C. Her optimism contrasts sharply with Geoffrey Hinton, who has argued the same technology poses existential risks that justify pausing development.
9. Why Did Mira Murati Sign a Gigawatt Compute Deal With Nvidia in 2026?
"This partnership accelerates our capacity to build AI that people can shape and make their own, as it shapes human potential in turn."Mira Murati, Axios, March 2026
The March 2026 Nvidia partnership expands what humans can do with AI, not what AI does to them, Murati said. The deal commits Thinking Machines to a gigawatt of Vera Rubin compute starting in 2027, plus an undisclosed Nvidia equity stake. It places the company on its biggest stage yet, partnered with the chipmaker powering nearly every frontier AI model.
What to Read Next
- Dario Amodei quotes on AI safety: compare Anthropic's safety-first corporate model with Murati's nonprofit-then-startup arc
- Yoshua Bengio's AI safety warnings: read his case that current training produces unpredictable systems against Murati's view that capability and safety align
- Yann LeCun quotes on LLM limits: LeCun's open-source advocacy compared with Murati's Thinking Machines approach
- Demis Hassabis on AGI and scientific discovery: DeepMind's AGI timeline next to Murati's "far from aligned" assessment
- All AI quotes: the full Aiifi collection of AI leader perspectives
