Renata Barreto, JD & PhD

Senior Research Scientist, eBay

prof_pic.jpg

Senior Research Scientist, eBay

Renata Barreto is a Senior Research Scientist at eBay, where she works on AI evaluations and AI risk governance. She holds a JD/PhD from UC Berkeley, where she focused on technical AI policy, model behavior, and the sociotechnical evaluation of AI systems.

Previously, she held research positions at Twitter, Meta, and Spotify across Responsible AI and Trust & Safety teams.

Her current research studies the societal impacts of agentic and generative systems, building technical mitigations and the evidence base for the policies institutions actually adopt. Recent work examines turning agentic governance proposals into testable evaluations, the limits of prompts-as-policy under fine-tuning, and persuasive dark patterns in chatbots.

She moves easily between languages, human and programming: fluent in Spanish, with advanced proficiency in French and Portuguese, and equally at home in Python, SQL, and PyTorch.

news

Sep 06, 2026 Presenting Alignment Inertia at the Trust and Safety Research Conference at Stanford, Oct 1–2, 2026.

selected publications

  1. ATTRIB
    Alignment Inertia: Auditing the Durability of Training Data Influence Through Policy Override Resistance
    Renata Barreto, Markelle Roesti, and Mohammad Tahaei
    2026
    Under review, NeurIPS Workshop on Attributing Model Behavior at Scale (ATTRIB)
  2. TAI
    Objection Without Action: Conversational Dark Patterns in AI-Mediated Commerce
    Francesco Salvi, Alejandro Cuevas, Renata Barreto, and 4 more authors
    2026
    Under review, NeurIPS Workshop on Trustworthy AI