# Logan Graham

Research and product · Head of Frontier Red Team at Anthropic · joined 2022

A Rhodes Scholar and former special adviser to the UK Prime Minister. He leads Anthropic's Frontier Red Team.

Full profile: https://www.autumn.ai/resources/anthropic/people/logan-graham

**At Anthropic:**
- Nov 2022 – present: Head of the Frontier Red Team
Sources: [1](https://www.linkedin.com/in/logangraham)

**Career before Anthropic:**
- Jul 2020 – Jul 2022: 10 Downing Street, Special Adviser to the Prime Minister
- Apr 2019 – Jul 2019: X, the moonshot factory, AI Resident
Sources: [1](https://www.linkedin.com/in/logangraham)

**Education:**
- D.Phil (PhD), Machine Learning (Engineering Science), University of Oxford, 2020
- BA Honours, Economics, The University of British Columbia, 2015
- National University of Singapore, 2013
Sources: [1](https://www.linkedin.com/in/logangraham)

**Lesser-known facts:**
- Did a DPhil in machine learning at Oxford. [1](https://www.linkedin.com/in/logangraham)
- Was an AI resident at X, Google's moonshot factory. [1](https://www.linkedin.com/in/logangraham)
- Helped establish the UK research agency ARIA while a special adviser at 10 Downing Street from 2020 to 2022. [1](https://www.alphaxiv.org/@logan-graham)
- Told NBC News to plan for Mythos-level hacking capabilities becoming broadly available within six to 12 months, not only from US companies. [1](https://www.nbcnews.com/tech/security/anthropic-claude-mythos-ai-hackers-cybersecurity-vulnerabilities-rcna273673)
- Called a July 2026 incident shared by OpenAI the first true AI safety incident and told his team to "remember this moment." [1](https://x.com/logangraham/status/2079991846705721351)
- Offered founders of independent AI auditors advice, connections and maybe money, writing "Let a thousand METRs bloom." [1](https://x.com/logangraham/status/2098860520086667388)
- Testified at a New York City Council AI hearing in October 2026, saying Anthropic welcomes "smart" regulation. [1](https://council.nyc.gov/press/2026/10/05/3278/) [2](https://www.krqe.com/news/business/ap-ai-industry-insiders-voice-alarms-to-nyc-council-as-local-governments-take-up-safety-concerns/amp/)
- Noted pandemics have cost on the order of $10 trillion since 2020, while only about $100 billion has gone to preventing the next one. [1](https://x.com/logangraham/status/2104411404547612696)

**Interviews and talks:**
- [Great Stagnation Summit panel "Unlocking the Potential" with Matt Clifford and Saffron Huang (2023)](https://www.youtube.com/watch?v=DPfzXBz8k7o)
- [Fox Business Mornings with Maria on Mythos and cybersecurity (2026)](https://www.youtube.com/watch?v=yI-Eovv2aSE)
- [Podcast: Why cyber defense can scale faster than AI attackers](https://podcasts.apple.com/us/podcast/why-cyber-defense-can-scale-faster-than-ai-attackers/id1891088763?i=1000796782049)

**Known for:**
- Constitutional classifiers: Defending against universal jailbreaks across thousands of hours of red teaming, 2024
- Sleeper agents: Training deceptive llms that persist through safety training, 2024
- MultiVerse: Causal Reasoning using Importance Sampling in Probabilistic Programming, 2020
- Toward trustworthy AI development: mechanisms for supporting verifiable claims, 2020
- The Future of Healthcare: Computerisation, Automation, and General Practice Services, 2020
- Artificial intelligence in hiring: Assessing impacts on equality, 2020
Sources: [1](https://www.congress.gov/119/meeting/house/118723/witnesses/HHRG-119-HM08-Bio-GrahamPhDL-20251217.pdf)

Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-10-18. Everyone at Anthropic: https://www.autumn.ai/resources/anthropic/people