A biosecurity nonprofit working to protect humanity against catastrophic pandemics through AI risk evaluation, pathogen-agnostic early warning surveillance, and DNA synthesis screening.
A biosecurity nonprofit working to protect humanity against catastrophic pandemics through AI risk evaluation, pathogen-agnostic early warning surveillance, and DNA synthesis screening.
People
Updated 04/02/26AIxBio Research Scientist
Funding Details
Updated 04/02/26- Annual Budget
- $7,324,329
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $15,079,587
Org Details
Updated 04/02/26SecureBio is a 501(c)(3) biosecurity nonprofit dedicated to securing the future against catastrophic pandemics, whether natural or engineered. Founded in 2022 by Kevin Esvelt, an MIT professor who invented CRISPR-based gene drive technology, the organization is headquartered at One Broadway, 13th Floor, Cambridge, MA 02142. SecureBio's work is organized around three strategic pillars: Delay, Detect, and Defend. The Delay pillar focuses on restricting dangerous pathogen genome proliferation and deterring harmful actors, primarily through the SecureDNA initiative which provides fast, comprehensive, and privacy-preserving DNA synthesis screening software. SecureDNA has since spun out into an independent Swiss foundation. The Detect pillar centers on the Nucleic Acid Observatory (NAO), led by Jeff Kaufman, which pioneers pathogen-agnostic early warning of future pandemics through deep, untargeted metagenomic sequencing of wastewater. By 2025, the NAO had scaled to 31 sampling sites across 19 US cities, achieving early detections of measles in Hawaii and West Nile virus in Missouri. The Defend pillar has explored germicidal ultraviolet light technology and pandemic-proof personal protective equipment, though this work is currently less active. A major and growing focus for SecureBio is the intersection of AI and biological risk. The AI & Biotechnology Risks team, led by Seth Donoughe, develops benchmarks and evaluation tools to assess whether frontier AI systems could lower barriers to bioweapons development. Their Virology Capabilities Test (VCT) has been adopted by all major frontier AI labs including Anthropic, OpenAI, Google DeepMind, Meta, and xAI for pre-release model evaluations. The team also develops mitigations including pretraining data filtering topic lists and training datasets deployed into frontier labs' safety pipelines. SecureBio is a member of the NIST US AI Safety Consortium and has engaged with the European AI Office. In September 2024, Kevin Esvelt stepped back from the Board to focus on his MIT professorship, and Benjamin Mueller (formerly COO) became Executive Director and Chairman of the Board. Three new board members joined: Christine Parthemore (CEO, Council on Strategic Risks), Michael Specter (Staff Writer, The New Yorker), and Liv Boeree (science communicator and philanthropist). The organization tripled its headcount in 2025 and by early 2026 has approximately 41 staff members across its AI, Detection, and Operations divisions. SecureBio has received significant funding from Open Philanthropy/Coefficient Giving (approximately $9.4 million across multiple grants), the Survival and Flourishing Fund ($1.556 million across three rounds), Sentinel Bio ($600,000), and the FTX Future Fund ($1.2 million). The organization is recommended by Founders Pledge and listed on Giving What We Can.
Theory of Change
Updated 04/02/26SecureBio's theory of change rests on the premise that advances in biotechnology and AI are making it increasingly feasible for malicious actors to create or acquire pandemic-capable pathogens, potentially leading to catastrophic or existential-level biological events. Their Delay-Detect-Defend framework addresses this by: (1) Delaying threats through DNA synthesis screening that prevents would-be bad actors from obtaining dangerous biological material, (2) Detecting novel pathogens early through pathogen-agnostic wastewater surveillance before they can spread to more than 1% of the population, giving civilization time to respond, and (3) Defending through physical and institutional measures that work regardless of the specific pathogen. On the AI front, SecureBio believes that rigorous evaluation of frontier AI models for biosecurity-relevant capabilities, combined with targeted mitigations like pretraining data filtering, can reduce the risk that AI systems become force multipliers for bioweapons development. By making these evaluations standard practice across all frontier labs, they aim to create an industry-wide norm of biosecurity responsibility.
Grants Received
Updated 04/02/26Projects– no linked projects
Updated 04/02/26Discussion
No comments yet. Be the first to share your thoughts.