Back to Headlines
Tech
Jun 05, 2026
Analyzed by GPT OSS 120B

Mira Murati Returns to Spotlight, Unveils ‘Interaction Models’ and Warns of Governance Gaps

AI Summary
Mira Murati, former OpenAI CTO and now CEO of Thinking Machines Lab, gave her first extensive interview in 18 months, previewing the startup’s new “interaction models” and highlighting governance challenges in the fast‑moving AI sector.

Mira Murati’s First Major Media Appearance in 18 months

In a Bloomberg interview in San Francisco, Mira Murati stepped back onto the public stage after a prolonged period of quiet. The former OpenAI CTO, now leading Thinking Machines Lab, used the conversation to signal the company’s re‑emergence and to remind the market that it remains a contender in the AI talent and funding race.

Introducing “Interaction Models”: Real‑Time Multimodal AI

Murati previewed the startup’s flagship concept called “interaction models”. Unlike the turn‑based, prompt‑and‑response paradigm that dominates most AI products, these models process continuous streams of audio, text, and video in 200‑millisecond intervals, aiming to capture the nuances of human conversation—interruptions, mid‑thought corrections, and pauses.

  • Product in early testing: Tinker, an API for fine‑tuning open‑source models.
  • Development timeline: ~1.5 years of background work (fundraising, hiring, product build).
  • Talent compensation trends referenced: nine‑figure packages becoming standard in the AI talent war.

Governance Concerns Amid AI Talent Wars

Murati shifted the discussion toward a broader industry issue: the concentration of consequential decisions in a handful of leaders. She warned that “good people make bad calls” and that the sector lacks robust structural checks, echoing concerns about the 2023 OpenAI board upheaval where she served as interim CEO for a five‑day “blip.”

When pressed about recent departures of high‑profile researchers from Thinking Machines, Murati framed turnover as a natural compression of years of organizational volatility into months, noting that compensation alone does not explain the movement.

What’s Next for Thinking Machines and the Wider AI Landscape

Murati declined to set a launch date for the interaction models, describing them as a “first step.” She emphasized that the current period will shape whether AI leads to dystopia or utopia, and that premature relinquishment of human oversight could steer outcomes “not better.”

Looking ahead, Murati’s measured tone suggests Thinking Machines will continue to iterate on real‑time multimodal interfaces while advocating for stronger governance frameworks across the industry.