Lilian Weng, co-founder of Thinking Machines, left the startup citing health reasons. Later, she rejoined OpenAI, where she had previously served as VP of AI Safety Research. The news, reported by TechCrunch , caught the eye of watchers across the global AI scene.
Therefore, the move has a dual significance. On one hand, it signals a potential fragility in next-gen AI startups, where founders face high operational pressures. On the other, it confirms OpenAI's magnetic pull as a hub for top researchers. Furthermore, the return of such a prominent figure in the safety area reinforces OpenAI's narrative as a responsible player in AI governance.
In short, for marketing and digital managers of Italian companies, this episode offers a useful insight: the consolidation of AI talent in major players is reshaping competitive balances. We at SHM Studio we monitor these movements to guide our clients' tech choices in the field of applied artificial intelligence . So, figuring out who leads AI research today means anticipating tomorrow's market directions.
The timeline: from Thinking Machines to OpenAI, a bumpy ride
Lilian Weng is a leading figure in AI research. She spent years at OpenAI as VP of AI Safety Research, building a solid reputation in model safety. Later, she co-founded Thinking Machines, an AI startup that had attracted international attention.
However, the journey hasn't been without bumps. According to reports from TechCrunch , Weng left Thinking Machines citing health reasons. Shortly after, she rejoined OpenAI. The speed of her return fueled speculation about the startup's future and OpenAI's specific weight as an employer in the AI sector.
Therefore, the sequence of events deserves a careful read. This isn't just a simple switch of sides. Instead, it's a signal about the internal dynamics of the global AI ecosystem.
Winners and losers: who's winning and who's losing in this setup
OpenAI is the most obvious winner. Re-acquiring a researcher of Weng's caliber in the AI safety area strengthens its position precisely at a time when global regulatory pressure is increasing. In fact, AI governance has become a competitive factor, not just an ethical one.
Thinking Machines, on the other hand, suffers a significant blow. Losing a co-founder is always a structural challenge. Beyond that, the public narrative — health reasons, then a quick return to a competitor — creates uncertainty around the vision and stability of the original team.
So, who really loses? The independent AI startup ecosystem. Every time top talent returns to the big players, the competitive diversity of the sector is reduced. Consequently, the oligopoly of a few players with unlimited resources is further consolidated.
Why AI safety matters so much in 2026: why this role counts
AI safety is no longer an academic niche. By 2026, it has become an operational priority for any company developing or integrating advanced language models. According to McKinsey , AI systems governance is among the top three concerns for C-level executives in organizations adopting generative technologies.
Weng directly contributed to building OpenAI's internal security frameworks. Furthermore, her work has influenced key technical literature in the field. Her return, therefore, is not just symbolic. It indicates that OpenAI intends to strengthen its research infrastructure precisely in the most sensitive area.
For Italian companies considering the adoption of solutions based on OpenAI models, this is a positive sign. In particular, it suggests that the provider continues to invest in the reliability and control of its systems. We at SHM Studio we consider the robustness of the safety layer as a relevant criterion when selecting AI tech partners for our clients.
SHM Studio's take: what this story shows about the AI market
This episode is a textbook case of talent gravitation within the AI ecosystem. Big players — OpenAI, Google DeepMind, Anthropic — exert a pull that startups struggle to counter. It's not just about compensation. Instead, factors like research scale, access to computational resources, and institutional visibility come into play.
Similarly, a recurring pattern is observed: AI startup founders coming from big labs tend to return to their original contexts after periods of operational stress. This doesn't mean startups are doomed to fail. However, it signals that the model
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
