# Mira Murati

> 1988– · Engineer; Thinking Machines Lab founder and CEO; former OpenAI CTO
>
> **Recorded contribution:** OpenAI product and research leadership through 2024; founded Thinking Machines Lab and serves as CEO

## How to use this dossier

Read for a causal chain, not a hero story: inherited problem → contribution → mechanism → downstream capability → limit. Then close the page and complete the reconstruction exercise from memory.

## 1. Historical orientation

Albanian-American engineer Mira Murati joined OpenAI in 2018 and became chief technology officer in 2022 after leading product and partnership functions. She helped oversee the organizational path to releases including ChatGPT and GPT-4 and briefly served as interim CEO in November 2023. Murati left OpenAI in September 2024 and founded Thinking Machines Lab in 2025, whose site listed her as CEO at verification time; the registry label 'CTO of OpenAI' is historical, not current. Murati's dossier exposes an often hidden layer of computing history: productization is a sequence of technical commitments about access, latency, interfaces, safety filters, monitoring, and rollback. Leadership matters when it creates a process in which research uncertainty is translated into release evidence and accountable operational decisions.

## 2. The problem inherited

Moving a frontier model from research into public use required product decisions, infrastructure, multimodal interfaces, safety evaluation, partnerships, and feedback loops operating under high uncertainty.

## 3. The central contribution

Murati supplied senior product and technical leadership during OpenAI's transition to mass-market generative-AI products and later founded a new AI laboratory.

## 4. Reconstruct the mechanism

1. Select a research capability and define the user workflow and prohibited outcomes around it.
2. Coordinate model, product, infrastructure, policy, red-team, and partnership readiness.
3. Stage access, monitor real interactions, and route failures into model and product evaluation.
4. Establish organizational ownership and stop or rollback authority as usage and capability change.

## 5. What changed downstream

- OpenAI's launches made conversational and multimodal model interfaces widely used.
- Her departure and new lab demonstrate the mobility of frontier-AI teams and the centrality of institutional design.
- Her tenure illustrates how launch leadership became historically important as general-purpose models moved from laboratory demonstrations to tools used by hundreds of millions of people.

## 6. Attribution, limits, and uncertainty

- Murati did not individually create ChatGPT or GPT-4; public leadership descriptions cannot resolve team-level technical attribution.
- Thinking Machines Lab is young and roles are volatile. Company and media accounts should not be converted into unverified technical claims.
- Public accounts cannot determine which individual made each technical decision, and both OpenAI and Thinking Machines have incentives to frame leadership; titles must be dated and independently checked.

## 7. Reconstruction lab

Write a release dossier for a multimodal assistant: intended use, evidence thresholds, red-team cases, access stages, monitoring, incident owner, and one irreversible-harm stop condition. Run a tabletop incident in which the model passes benchmarks but causes a novel real-world harm, and document who can pause access before organizational incentives override evidence.

## 8. Evidence trail

- [OpenAI leadership team update](https://openai.com/index/leadership-team-update/) — OpenAI
- [Mira Murati leaves OpenAI](https://apnews.com/article/df75217584696b442935dbccc9b0347d) — Associated Press
- [Thinking Machines Lab founding team](https://thinkingmachines.ai/) — Thinking Machines Lab

---

*Research checked 2026-08-09. Dates, roles, and claims about living people are historical snapshots. Linked sources remain the authority; this dossier is original instructional synthesis.*