Connor Leahy is a German-American artificial intelligence researcher and entrepreneur known for his work on large language models and his advocacy on the risks of advanced AI. He co-founded EleutherAI, a collective that produced early open-source language models, and later founded the AI safety research company Conjecture, which he led as CEO until 2026. As of 2026, he serves as US Director of ControlAI, a nonprofit focused on preventing extinction risk from artificial superintelligence.
Leahy has been a prominent voice warning about the potential for human extinction from advanced AI. He has testified before the House of Commons of Canada and participated in the inaugural AI Safety Summit in 2023. His public statements often emphasize the unpredictability of powerful AI systems and the need for government oversight.
Early Life and Education
Leahy grew up in Germany and later moved to the United States. He studied at a university where he began experimenting with Machine learning models in his spare time. While still a student in 2019, he reverse-engineered OpenAI's GPT-2 in his dormitory room, at a time when the full model had been initially withheld by the organization. This early project sparked his interest in the inner workings of large language models and their potential societal impact.
Career and Research
In July 2020, Leahy co-founded EleutherAI together with Sid Black and Leo Gao, with an initial focus on replicating GPT-3 in an open-source setting. During his time as co-lead, EleutherAI produced some of the earliest open-source large language models and datasets. These included The Pile in December 2020, GPT-Neo in March 2021, GPT-J-6B in June 2021, and GPT-NeoX-20B in February 2022. At the time of its release, GPT-J-6B was the largest open-source GPT-3-style language model in the world, a significant milestone for the AI research community.
In March 2022, Leahy co-founded Conjecture, an AI safety research company that he led as CEO. The company's stated mission was to scale applied AI alignment research, aiming to address the challenge of ensuring that powerful AI systems act in accordance with human intentions. Leahy has expressed skepticism about reinforcement learning from human feedback as a complete solution to the alignment problem. He has described powerful AI systems as "alien" in their reasoning, noting that techniques like RLHF may only mask underlying non-human thought processes. He has used the metaphor of a "smiley face mask" that can slip under unexpected prompts, revealing what he calls a "massive underbelly of insanity."
Advocacy and Public Engagement
In March 2023, Leahy was one of the signatories of an open letter from the Future of Life Institute calling for "all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4." The letter drew widespread attention and sparked debate about the pace of AI development. In May 2023, he was among the signatories of the Statement on AI Risk, a one-sentence declaration that reads: "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." The statement was signed by hundreds of notable figures, including Nobel laureates, AI researchers, and CEOs of major AI companies such as OpenAI, Anthropic, and Google DeepMind.
In November 2023, Leahy was invited to speak at the inaugural AI Safety Summit held in the United Kingdom. He expressed concern that the summit would fail to adequately address the risks from AI systems surpassing human capabilities, arguing that if such systems are built without proper control, "the future belongs to them, not us." He worked with the nonprofit ControlAI to advocate for governments to implement a pause on the development of artificial general intelligence, and criticized major AI companies for attempting to evade government oversight and regulation.
Later Career
In March 2026, Leahy announced in a retrospective that the Conjecture project was drawing to a close, and that he was joining ControlAI as its US Director, a role in which he had previously served as an advisor. His work at ControlAI focuses on policy advocacy and public education regarding the risks of artificial superintelligence. He continues to write and speak publicly about AI safety, maintaining a Substack newsletter titled "Expedition to the Far Lands."
See Also
- Llion Jones
- Jakob Uszkoreit
- Lukasz Kaiser
- Niki Parmar
- Mark Chen
- Brad Lightcap
- Jacob Steinhardt
- David Kaplan
- Jack Clark
- David Luan
- Chen Wu
- Ashish Kumar
- Karen Simonyan
- Koray Kavukcuoglu
- Michael Jordan
- Anima Anandkumar
- Samy Bengio
- Joshua Tenenbaum
- Brendan Lake
- Melanie Mitchell