PeopleofSuperintelligence

Researcher

Dawn Song

Dawn Song is a UC Berkeley computer scientist whose research connects AI with computer security, privacy and program synthesis. She earned degrees at Tsinghua, Carnegie Mellon and UC Berkeley, then held an assistant professorship at Carnegie Mellon from 2002 to 2007 before joining Berkeley. The MacArthur Foundation named her a 2010 Fellow, and her 2017 neural programming paper won an ICLR Best Paper Award. She later founded Oasis Labs and co-authored DecodingTrust, a NeurIPS 2023 Outstanding Paper that assessed GPT trustworthiness. Today, she co-directs Berkeley's Center for Responsible Decentralized Intelligence, contributes to Vellum's AI-aided reasoning work, and is an ACM and IEEE Fellow and a member of the American Academy of Arts and Sciences.

Works
602
Citations
71,236
h-index
116

Publication numbers from OpenAlex, as of 14 October 2026.

Photo of Dawn Song

Timeline

  1. 1996–2007

    Training across three universities

    She earned a B.S. from Tsinghua University in 1996, an M.S. from Carnegie Mellon University in 1999, and a Ph.D. from UC Berkeley in 2002. From 2002 to 2007, she was an assistant professor at Carnegie Mellon before joining Berkeley's faculty.

  2. 2010

    A MacArthur Fellowship

    The MacArthur Foundation named her a 2010 Fellow for work that analyzed vulnerable software and developed defenses against computer attacks.

  3. 2017

    Teaching programs to recurse

    With Jonathon Cai and Richard Shin, she published an ICLR 2017 paper that added recursion to neural programming architectures. It evaluated grade-school addition, bubble sort, topological sort and quicksort, and received a Best Paper Award.

  4. 2018

    A privacy platform takes shape

    She founded Oasis Labs in 2018 around a blockchain vision for privacy-preserving data use, with people controlling their own data.

  5. 2023

    Testing trust in GPT

    She co-authored DecodingTrust, a NeurIPS 2023 Outstanding Paper in the Datasets and Benchmarks track. The benchmark assessed GPT-4 and GPT-3.5 for toxicity, stereotype bias, adversarial robustness, privacy, ethics and fairness.

  6. 2025

    Reasoning with theorem provers

    She joined a Berkeley and UT Austin collaboration on Vellum, which received an AI for Math Fund grant in 2025. The open-source framework connects large language models with interactive theorem-provers and automated solvers for AI-aided formal reasoning in mathematics and computer science.

  7. As of October 2026

    Guiding responsible AI

    She is a UC Berkeley professor and co-director of the Center for Responsible Decentralized Intelligence, with research in AI safety, security, deep learning and agentic AI. She is an ACM and IEEE Fellow and was elected to the American Academy of Arts and Sciences in 2025.

Sources and credits

Photo: Xuthoria, via Wikimedia Commons. License: CC BY-SA 4.0. Source: commons.wikimedia.org. Cropped to a square and resized. Ask us to remove or replace this photo.

  1. Dawn Song - MacArthur Foundation macfound.org. Checked 14 October 2026.
  2. Dawn Song | USENIX usenix.org. Checked 14 October 2026.
  3. Richard Shin rshin.github.io. Checked 14 October 2026.
  4. [1704.06611] Making Neural Programming Architectures Generalize via Recursion arxiv.org. Checked 14 October 2026.
  5. The Genesis of the Oasis Foundation oasis.net. Checked 14 October 2026.
  6. Announcing the NeurIPS 2023 Paper Awards blog.neurips.cc. Checked 14 October 2026.
  7. UC Berkeley receives AI for Math Fund grants for Vellum and LeanTutor projects cdss.berkeley.edu. Checked 14 October 2026.
  8. Dawn Song | EECS at UC Berkeley www2.eecs.berkeley.edu. Checked 14 October 2026.