Thomas Yang
Role: Founder, Phi Lab Foundation
Headline: Founder of Phi Lab Foundation, leading the Foundation’s research program in AI security and trustworthy AI.
Research interests: AI security and trustworthy AI, with a focus on watermarking and provenance of model-generated content (text, image, audio, DNA), detection of synthetic media, adversarial robustness, and red-teaming of foundation models. Broader interests in cybersecurity, model misuse and alignment, and information integrity.
Profile summary: Thomas Yang founded Phi Lab Foundation, a 501(c)(3) non-profit research organization, and leads its research team. He sets the Foundation’s research agenda, supervises its research staff and collaborators, and is responsible for the design and execution of its studies in content provenance and AI security. He also leads the Foundation’s external academic collaborations and its organization of peer-reviewed academic venues.
Research leadership:
- Directs the Foundation’s research program in watermarking, provenance, and synthetic media detection, and is accountable for its scientific direction and standards.
- Supervises the Foundation’s interdisciplinary research team and coordinates collaborations with university-based researchers.
- Lead organizer of SIGNAL @ WACV 2027, a peer-reviewed academic workshop on source-integrated generative-content attribution and labeling, held at IEEE/CVF WACV 2027.
- Responsible for research outputs, publication, and the Foundation’s open-source release of benchmarks and evaluation protocols.
Selected work:
- Watermarking and provenance methods spanning multiple carriers, including text, image, audio, and DNA sequences.
- Evaluation protocols and benchmarks for robustness of provenance signals under real-world transformations and adversarial removal.
- Red-teaming and adversarial robustness studies of foundation models.
Contact: [email protected]
LinkedIn: Profile