PeopleofSuperintelligence

Lab leader

Marina Favaro

Marina Favaro is an AI policy leader whose work connects nuclear security, model safety and public input to AI governance. She managed BASIC's Emerging Technologies programme, researched nuclear stability at King's College London and IFSH, and joined Anthropic's policy team in 2023. At Anthropic, she supported research on a public-input AI constitution, coauthored a study of deceptive language models and oversaw national-security partnerships. In 2026, she coauthored research on recursive self-improvement and public measures of AI development; as of October 2026, she leads The Anthropic Institute and holds a master's degree in international relations and politics from the University of Cambridge.

Now at
Anthropic
Badges

Timeline

  1. 2020–2021

    BASIC's Emerging Technologies programme

    From 2020 to 2021, she manages BASIC's Emerging Technologies research programme. Before this role, she works as an analyst at RAND Europe, where her research focuses on space security, cybersecurity, defence innovation and the impact of emerging technologies on society.

  2. 2021

    A nuclear stability study

    In 2021, she leads a King's College London Centre for Science and Security Studies study of how emerging technologies affect nuclear stability. The study uses machine-learning clustering to group ten technologies into four clusters by their effects, impact and feasibility.

  3. 2022

    Negative Multiplicity at IFSH

    In 2022, she is a Visiting Research Fellow in IFSH's Arms Control and Emerging Technologies project and coauthors Research Report 010, Negative Multiplicity, with Neil Renic and Ulrich Kühn. The report forecasts how emerging technologies may affect international stability and human security.

  4. March 2023

    Anthropic policy and nuclear security

    In March 2023, she appears as Anthropic's Senior Policy Analyst at an arms-control session on AI's overlap with the nuclear domain and the private sector's role in standards and norms.

  5. October 2023

    Collective Constitutional AI

    In October 2023, she supports Collective Constitutional AI, Anthropic and the Collective Intelligence Project's study that involves about 1,000 Americans in drafting an AI constitution. Anthropic thanks her for support and guidance on the study.

  6. 2024

    Deceptive backdoors in Sleeper Agents

    In 2024, she coauthors Sleeper Agents, a study of deceptive language models and their backdoors. The authors find that backdoor behavior can persist through supervised fine-tuning, reinforcement learning and adversarial training.

  7. 2025

    National security partnerships at Anthropic

    In October 2025, she oversees Anthropic's National Security Policy & Partnerships and describes deploying Claude in a Top Secret environment for NNSA testing of nuclear risks. In December 2025, she joins a Bay Area Council panel on AI-enabled cyber threats, government-industry collaboration and technical standards.

  8. As of October 2026

    Leading The Anthropic Institute

    In June 2026, she coauthors When AI builds itself with Jack Clark, using public benchmarks and Anthropic's internal data to assess progress toward recursive self-improvement and its implications. In September 2026, she coauthors public measures of AI's share of R&D work, human oversight of AI agents and compute allocation to safety, with Anthropic results reported as of August 2026. As of October 2026, she leads The Anthropic Institute at Anthropic and holds a master's degree in international relations and politics from the University of Cambridge.

Career

  1. Since 2023AnthropicSenior Policy Analyst
  2. 2022Institute for Peace Research and Security Policy at the University of HamburgVisiting Research Fellow
  3. 2021King's College LondonConsultant
  4. 2020–2021British American Security Information CouncilManager
  5. RAND CorporationAnalyst

Education

  1. University of CambridgeMaster's degree

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Marina Favaro. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. 16% of AI career moves in the 2020s went from a university to a companyIncludes Marina Favaro. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  3. Of 216 AI researchers who entered public life, 83 held advisory seats and 58 took public rolesIncludes Marina Favaro. Of 216 AI researchers who entered public life, 83 held advisory seats and 58 took public roles. Another 97 people moved from public careers into AI organizations or research, with 62% of dated records starting from 2020 on.
  4. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Anthropic. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Marina Favaro yet. Send us a free-licensed photo.

  1. Marina Favaro | European Leadership Network europeanleadershipnetwork.org. Checked 21 October 2026.
  2. Emerging technologies could risk catastrophic nuclear escalation | King's College London kcl.ac.uk. Checked 21 October 2026.
  3. Negative Multiplicity: Forecasting the Future Impact of Emerging Technologies on International Stability and Human Security | University of Copenhagen Research Portal researchprofiles.ku.dk. Checked 21 October 2026.
  4. Marina Favaro | Arms Control and Emerging Technologies hamburg-armscontrol.de. Checked 21 October 2026.
  5. Young Women and Next Generation Initiative – The EU Non-Proliferation Consortium nonproliferation.eu. Checked 21 October 2026.
  6. Collective Constitutional AI: Aligning a language model with public input anthropic.com. Checked 21 October 2026.
  7. Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training arxiv.org. Checked 21 October 2026.
  8. Anthropic Has a Plan to Keep Its AI From Building a Nuclear Weapon. Will It Work? | WIRED wired.com. Checked 21 October 2026.
  9. Council Builds Global Partnerships to Advance AI Leadership | Bay Area Council bayareacouncil.org. Checked 21 October 2026.
  10. When AI builds itself anthropic.com. Checked 21 October 2026.
  11. Anthropic: IA em breve poderá se aperfeiçoar sem intervenção humana | CNN Brasil cnnbrasil.com.br. Checked 21 October 2026.
  12. Measurements for understanding the pace of AI development inside frontier labs anthropic.com. Checked 21 October 2026.