Dr. Rebecca Portnoff

Standards & Policy

A central focus of my work has been bringing together technical research, operational experience building and scaling AI, and deep expertise in child sexual exploitation and abuse to shape how AI and other technologies are governed in practice. My work spans originating Safety-by-Design approaches and driving industry adoption, contributing technical and issue-area expertise to standards and regulatory frameworks, co-developing guidance with public institutions, and advising governments on emerging AI and child-safety risks.

Technical Standards & Governance

I translate technical research, operational experience, and child-safety expertise into standards, recommended practices, and governance frameworks that shape how AI systems are designed and governed. Selected contributions include:

Safety-by-Design for Generative AI

Conceived, established, and led Thorn’s Generative AI Safety-by-Design initiative, developing concrete principles and mitigations for preventing generative AI-facilitated child sexual abuse. Built the multi-stakeholder initiative and secured public commitments to the principles from more than ten leading AI companies, including Google, Anthropic and OpenAI.

UK AI Security Institute

Co-authored the recommended practice jointly developed by Thorn and the UK AI Security Institute for preventing AI-generated CSEA, providing practical guidance for developers and other organizations implementing preventative measures.

IEEE

The Safety-by-Design for Generative AI work I led became the basis for IEEE P3462, a recommended practice formalizing Safety-by-Design approaches for child safety in generative AI.

NIST

Provided technical and issue-area expertise at the intersection of AI governance and child sexual exploitation and abuse across multiple NIST standards and guidance efforts, including invited expert review and contributions to NIST AI 100-4 and the AI Standards Zero Drafts initiative.

EU AI Act

Selected to participate in the multi-stakeholder process developing the EU AI Act Code of Practice for general-purpose AI, contributing technical and child-safety expertise to how CSEA risks and mitigations were addressed within systemic-risk and compliance frameworks.

Australia’s eSafety Commissioner

Contributed technical and child-safety expertise to Australia’s eSafety Commissioner’s Generative AI Position Statement, which addresses CSEA and other emerging generative AI harms and sets out Safety by Design measures across the AI lifecycle.

Government Briefings & Global Convening

I regularly brief governments and public institutions in the United States and internationally on AI safety, technology governance, and child-safety risks. Selected engagements include:

Executive Office of the President & NIST

Briefed the Executive Office of the President and NIST on Safety-by-Design for Generative AI, joining civil-society and academic leaders to discuss approaches for preventing AI-facilitated child sexual exploitation and abuse.

White House

Participated as an expert in a White House roundtable on AI safety and technical approaches to preventing image-based sexual abuse.

International Network of AI Safety Institutes

Provided expert guidance on AI safety and child sexual exploitation and abuse at the inaugural convening of the International Network of AI Safety Institutes.

Australia’s eSafety Commissioner

Contributed expertise on Safety-by-Design, emerging generative AI harms, and online-safety interventions at eSafety’s Generative AI and Online Safety Roundtable, held at the Australian High Commission in London.

Internet Watch Foundation & UK Government

Presented technical and child-safety perspectives on preventing AI-generated child sexual exploitation and abuse at an Internet Watch Foundation roundtable with the UK Home Office, GCHQ, Ofcom, the National Crime Agency, and other stakeholders.

Policy Submissions & Public Comments

I contribute technical and child-safety expertise to public consultations shaping AI policy, standards, and regulation in the United States and internationally. Selected submissions and public comments include:

  • NIST RFI: Implementing the AI Executive Order
    National Institute of Standards and Technology (NIST)

    February 2024

    Submitted joint comments from Thorn and All Tech Is Human providing technical recommendations for preventing generative AI-facilitated child sexual exploitation and abuse, including Safety-by-Design interventions across the AI lifecycle.

  • AI Action Plan RFI
    White House Office of Science and Technology Policy (OSTP)

    March 2025

    Submitted recommendations for the U.S. AI Action Plan focused on preventing generative AI-facilitated child sexual exploitation and abuse, including federal investment in training-data safeguards, model assessment, protections against adversarial fine-tuning, and oversight of model-hosting platforms.

  • Scientific Contribution on Generative AI Risks to Minors
    French G7 Presidency / Ambassador for Digital Affairs and AI

    May 2026

    Submitted scientific and technical input on generative AI-facilitated sexual harms against minors, identifying mechanisms driving AI-enabled CSEA and recommending Safety-by-Design interventions across the AI lifecycle, including training-data filtering, provenance, model auditing, red teaming, protections against harmful fine-tuning, moderation, transparency, and cross-industry signal sharing.

  • Public Consultation on AI Risk Thresholds
    Organisation for Economic Co-operation and Development (OECD)

    September 2024

    Submitted technical input demonstrating limitations of compute-based AI risk thresholds through the lens of CSEA, and recommending capability- and training-data-based approaches that account for marginal risk, harmful training data, and domain-specific evaluation.

  • NIST Zero Draft for Documentation of AI Datasets and Models
    National Institute of Standards and Technology (NIST)

    October 2025

    Submitted technical recommendations for NIST’s proposed AI documentation standard, bringing child-safety and CSEA considerations into dataset provenance and evaluation, model assessment, and governance requirements. The submission recommends, among other things, documenting efforts to identify illegal material such as CSAM in training data, assessing data minimization, disclosing third-party model evaluations, and documenting enforcement and auditing mechanisms.