Strategies for Responding to Exponential AI and Biotechnology Growth
November 13, 2025
Hopkins Bloomberg Center
555 Pennsylvania Avenue NW, Washington DC
Background: The ever-increasing acceleration of agentic artificial intelligence (AI) and biological design tools has transformed the technological landscape, enabling tremendous benefits and potential misuse that could massively impact national security and public health. Mitigating this risk will require collaboration across Government, Industry, and Academia with both technical and policy focuses. Significant effort has already been made to raise awareness of this challenge but additional discussion is necessary to maintain pace with the speed of evolving technology.
Objective: Johns Hopkins and RTX BBN Technologies, with funding from a Johns Hopkins Nexus Award, are excited to host a workshop series to shape safeguards that can enable the community to understand potential risk and inform response on relevant timelines.
The November 13, 2025 workshop will seek to define a common problem statement that can be assessed from different perspectives in subsequent conversations. Specifically, we are looking for participants that are seeking to advance understanding of the following:
- Agentic AI capabilities
- AI-resilient safeguards,
- Biodesign tool capabilities,
- Biological mechanisms,
- Cyberbiosecurity,
- International efforts and harmonization,
- Robotic and cloud labs,
- Computational and physical safety protocols.
Run of Show: The morning will focus on level setting the communities’ understanding with keynote talks from experts and thought leaders; stay tuned for finalized agenda.
Afternoon sessions will facilitate structured discussions focused on various components of the problem with the objective of generating a comprehensive and agreed upon problem statement. These breakout conversations will discuss the following topics (topics may be augmented):
Capability Assessment Tools – Bio-design tools are evolving quickly; what experiments and protocols can be developed and routinely deploy to rapidly assess model uplift and autonomous execution? This group will NOT focus on knowledge benchmarks.
Cyberbiosecurity – What principles and practices can be adopted from cybersecurity practices that will work for the unique biological context?
Red teaming AI design tools – What is the art of the possible to identify and respond to potential misuse of novel design tools BEFORE they cause harm?
Outcome: This workshop will produce written findings and recommendations from each of the breakout groups for use in follow-on meetings.
Agenda
| Time | Description |
|---|---|
| 0830-0840 | TBD: Opening Remarks “Scene Setter” |
| 0840-0910 | Gene Godbold (SigSci): Influence of SOCs on bacterial pathogenesis |
| 0910-0940 | Coleman Breen (JHU): Statistical Best Practices for AI Capabilities Evaluations |
| 0940-0945 | BREAK |
| 0945-1010 | Stephanie Guerra (RAND): Emerging Best Practices for Developing Agent Capability Evaluations in Biosecurity |
| 1010-1030 | Allison Berke (RAND): Verifiable Audit Trails: A Proposal for Biodesign Logging |
| 1030-1050 | Jasper Gotting (SecureBio): SecureBio’s Biology Benchmarks Dashboard – Assessing LLM performance on biosecurity-relevant evaluations |
| 1050-1100 | BREAK |
| 1100-1120 | Moritz Hanke (JHU): The need for funding-stage pre-development risk-benefit review to reduce large-scale harm from biological AI models |
| 1120-1140 | Keltin Grimes (CMU): Machine unlearning techniques for biosecurity |
| 1140-1200 | Tessa Alexanian (IBBIS): Beyond Open vs. Secret: Managed Data Access in the Age of AI |
| 1200-1220 | Jake Beal (RTX BBN): An “Enforcement Ready” Biosecurity Screening Standard for Nucleic Acid Providers |
| 1220-1320 | LUNCH BREAK |
| 1320-1330 | Justin Taylor (Anthropic): Risk assessment and threat modeling to identify potential CBRN misuse vectors |
| 1330-1350 | Overview and Setup of Breakouts |
| 1350-1510 | Breakout work |
| 1510-1520 | BREAK |
| 1520-1600 | Summary and Next Steps |
| 1630-FINX | No Host Social at Penn Quarter Sports Tavern |
Abstract Submissions
We will select short talks based on 1-2 page abstracts (format of your choice). Abstracts will be compiled and distributed as a product of the meeting.
Submission site: Google Form
Deadline: October 20, 2025, AoE
Decisions: October 27, 2025
Slides due: November 10, 2025, AoE
Registration
We are at capacity for the venue. Virtual registration is still available.
Organizers
Joel Bader (JHU), Melissa Hopkins (JHU), Aaron Adler (BBN), Kemper Talley (BBN), Nic Roehner (BBN)