Database

No summary available yet.

Alexander Matt Turner, known online as TurnTrout, is a research scientist at Google DeepMind on the Scalable Alignment team, based in Berkeley, California. He earned his PhD in computer science from Oregon State University (2016–2022) under advisor Prasad Tadepalli, followed by a postdoc at UC Berkeley's Center for Human-Compatible AI (CHAI) from 2022 to 2023. His research spans several key areas of AI alignment: he developed Attainable Utility Preservation (AUP), a framework for low-impact AI; proved mathematically that optimal policies tend to seek power, published as a NeurIPS 2021 spotlight paper; co-developed shard theory (with Quintin Pope), a framework modeling AI and human values as situationally activated goal components; and pioneered activation engineering and steering vectors for controlling model behavior at inference time. At Google DeepMind, his current work includes consistency training to reduce sycophancy and jailbreaks in Gemini models. He is also a MATS mentor through his Team Shard program, supporting junior alignment researchers.

No summary available yet.

MSDS @ UCSD

Project Director at AI Plans, a platform leading research and events in AI alignment and safety.

No summary available yet.

Co-Founder and Growth lead at Heron AI Security Initiative, helping build a community of world-class cybersecurity experts working on high-leverage frontier AI security challenges.

Senior Lecturer in Information Technology at Deakin University whose research focuses on data mining, privacy-preserving analytics, anomaly detection and related areas, and who is listed as a member of the Australian Responsible Autonomous Agents Collective (ARAAC).

Jamie Bernardi (full name James Bernardi) is an AI safety researcher and policy professional based in London, UK. He studied Physical Natural Sciences (MSci Physics) at St John's College, Cambridge (2015-2019), then worked as a machine learning engineer at Audio Analytic (2019-2021), a sound recognition company later acquired by Meta. In 2021-2022 he worked as an independent researcher in collaboration with Oxford's Future of Humanity Institute, designing a pessimistic reinforcement learning algorithm intended to improve AI safety by having the system ask for human help when uncertain about potential harms. He co-founded BlueDot Impact in August 2022, an education charity running the AI Safety Fundamentals courses that have introduced thousands of ML engineers, policymakers, and students to AI safety and governance. After serving as a GovAI Winter Fellow and IAPS Fellow focused on AI policy and societal adaptation to advanced AI, he joined the UK AI Security Institute (AISI), where he works on strategic awareness. He has contributed to UK policymaking on AI regulation and international AI governance efforts.

Cambodian computer scientist and technology leader with around two decades of experience in data and artificial intelligence, former Google AI programme lead and current Head of Data and AI Ecosystems at the World Economic Forum, who co-founded and serves as Executive Director of AI Safety Asia while advising the UN Secretary-General’s High-Level Advisory Body on Artificial Intelligence.

No summary available yet.

I'd like to explore a research agenda at the intersection of time horizon model evaluation and control protocols.

No summary available yet.

No summary available yet.

Train LLMs to accurately & honestly report on their internal decision-making processes through real-time introspection

Effective altruist who works on worse-than-extinction outcomes from AI

Building infrastructure to give existential risk researchers superforecasting ability with minimal overhead

6-month salary for research into preventing steganography in interpretable representations using multiple agents

No summary available yet.

No summary available yet.

Open-Source Runtime Governance Architecture for Structural Alignment Drift in Long-Running AI Agents

4-month salary to research empirical and theoretical extensions of Cohen & Hutter’s pessimistic/conservative RL agent

No summary available yet.

4-month salary for conceptual/theoretical research towards perfect world-model interpretability

No summary available yet.

Nick Bostrom is a Swedish-born philosopher known for his work on existential risk and the long-term future of humanity. He was Professor of Philosophy at the University of Oxford and founding Director of the Future of Humanity Institute from 2005 until its closure in 2024, and is now the founder and principal researcher at the nonprofit Macrostrategy Research Initiative.

Educating the general public about AI and risks in most efficient ways and leveraging this to achieve good policy outcomes

No summary available yet.

Sofia Lanfri is a data scientist and research program manager with experience in geospatial and epidemiological analysis. She serves as Program Director at the Beneficial AI Foundation and coordinates BAIF’s involvement in Signal Shot and other AI safety projects.

Writer-researcher at Convergence Analysis

_̴̡̧̧̢̧̭̺̣͖̬̲̹̬̭̰̘̫͎̹͓̹̮͔̩̦̫̮̰͕̝̝̲̺͚̪̝̰̲̩̹̱̤͒̄̀̅̍͐̓̀͊̌̍̎̐̔͋̆̈́̽̀̓̀̀̾́̾̑͗́̉̃̎̾̊͑̆̓̀̒̎̈́͋͜ͅDorota Kontny

No summary available yet.

No summary available yet.

Orpheus Lummis is a software developer and community organizer based in Montréal, Canada, focused on advancing AI safety through convenings, coordination, and software. He is the founder of HΩ (Horizon Omega), a Canadian non-profit that runs the AI Safety Unconference series, the Guaranteed Safe AI Seminars, and a weekly AI Safety Events and Training newsletter. He has organized multiple editions of the AI Safety Unconference, including events co-located with NeurIPS and a fully virtual edition in 2024, and organized the Limits to Control Workshop in 2025. He has received funding from the Long-Term Future Fund for upskilling in AI techniques, deep reinforcement learning, and AI safety, and engaged with David Krueger's AI safety reading group at Mila. He is a software engineer generalist with specialties in distributed systems and machine learning, and was a core contributor to DefraDB, a decentralized user-centric database.

AI/AI safety Researcher.

5-month funding to continue upskilling in mechanistic interpretability post-SERI MATs, and to continue open projects

No summary available yet.

Director of the UC Berkeley Cybersecurity Clinic and lecturer at the School of Information, Elijah Baucom is a cybersecurity and privacy technologist, engineer, and activist focused on the intersection of technology, humanity, liberation, and political education; through the clinic he teaches and trains students to consult with and support social‑sector organizations that are particularly susceptible to ideologically motivated cyber attacks, and he is also the founder of Everyday Security, which provides cybersecurity and IT support to human rights and movement organizations.

Founder of PauseAI, CEO of Ontola.io

Technical Program Manager at the Technical Alignment Research Accelerator (TARA) and Master of AI student at UNSW, with a background including a Bachelor of Science in Mathematical Physics and a Diploma in Music from the University of Melbourne, completion of TARA v1 with a capstone project on causal interventions on a chess model, acceptance into the SPAR program, leadership roles in Effective Altruism University of Melbourne, and work on AI-focused education seminars.

Alex Flint is an independent AI alignment researcher and monastic. He completed his PhD with the Active Vision Lab at the University of Oxford (2008-2012) in robotics and computer vision on a Clarendon Fund scholarship, following undergraduate studies in computer science at the University of Adelaide. After his PhD he led computer vision engineering at Flyby Media in New York City, then co-founded Zippy.ai, a robotics startup focused on last-mile delivery, which was acquired by Cruise (GM's autonomous vehicle division) in 2018. He subsequently left industry to pursue independent AI alignment research and is affiliated with the Monastic Academy for the Preservation of Life on Earth in Vermont. His alignment research focuses on foundational theories of optimization, knowledge, and agency, including a widely cited post "The ground of optimization" selected as Best of LessWrong 2020. He has received funding from the Long-Term Future Fund for independent research and is active on LessWrong and the AI Alignment Forum.

Ben Hoffman has worked as a researcher for GiveWell and the Open Philanthropy Project and in financial services managing a risk analytics team, and he holds an MS in Mathematics and Statistics from Georgetown University and a BA from St John’s College.

Buck Shlegeris is the CEO of Redwood Research, a nonprofit AI safety and security lab, where he leads work on AI control and related empirical safety research; he previously worked on AI alignment and outreach at the Machine Intelligence Research Institute (MIRI).

No summary available yet.

No summary available yet.

Funding the open-source launch of a working claim-state system and the local firewall bridge that carries verification before voice into governed agent action.

Lorenzo is a full‑stack software engineer at Giving What We Can, supporting product development and maintaining the organisation’s technical infrastructure. Originally from Italy, he previously did small‑scale earning to give as a software developer in the Netherlands and, after winning a donor lottery in 2022, moved back to Italy to reflect on how to do more good.

Investigate humans’ lack of robust task alignment in amplification, and the implications for acceptability predicates