Technical Innovations for AI Policy

Washington, DC

May 30, 2025
June 1, 2025
Date Range

Overview

FAR.AI, in collaboration with FAI, CNAS, and RAND, organized the inaugural Technical Innovations for AI Policy Conference, which took place in Washington, D.C. from Saturday May 31 - Sunday June 1.

This conference aimed to inspire technical professionals, build strong networks within technical AI governance, and catalyze impactful work on pressing technical AI governance challenges.

Sessions covered topics including:

  • semiconductor supply chains and export controls,
  • hardware-enabled mechanisms and on-chip governance tools,
  • monitoring, reporting, and verification,
  • data center and model weight security,
  • defense and national security,
  • model evaluations,
  • energy and AI,
  • and more.
    In addition to talks and discussions, the agenda included substantial time for networking and collaboration.

The event was aimed at researchers or engineers with relevant expertise—especially in hardware—or  policymakers looking to learn more.

Financial support: This event is supported by FAR.AI’s general funds alongside generous contributions from Nevin Freeman and Brandon Goldman.

Planning to share or post about the event? You can find suggested language and graphics in the Media Kit.

Our speakers included:

  • Congressman Bill Foster
  • Saif Khan (Institute for Progress and Bipartisan Senior Advisor to the U.S. House Select Committee on the CCP)
  • Ben Buchanan (Johns Hopkins University)
  • Steve Kelly (Institute for Security and Technology)
  • Samuel Hammond (Foundation for American Innovation)
  • Mark Beall (AI Policy Network)
  • Lennart Heim (RAND Corporation)
  • Helen Toner (Center for Security and Emerging Technology)
  • Brad Carson (Americans for Responsible Innovation)
  • Miranda Bogen (Center for Democracy and Technology)
  • Tim Fist (Institute for Progress)
  • Arnab Datta (Institute for Progress)
  • Alex Bores (New York State Assembly)‍
  • Jenny Marron (Institute for AI Policy and Strategy)
  • Mary Phuong (Google DeepMind)
  • Sarah Cen (Stanford University)
  • Ben Bucknall (University of Oxford)
  • Charles Yang (Center for Industrial Strategy)
  • Lisa Thiergart (Institute for Security and Technology)
  • Irene Solaiman (Hugging Face)
  • Chris Byrd (RAND Corporation)
  • Abigail Hoskin (Horizon Institute for Public Service)
  • Kevin Wei (RAND Corporation)
  • Adam Gleave (FAR.AI)

Technical Innovations for AI Policy sessions

Verifiable Compute & Building Trust in Agentic Supply Chains

Tina Morrison

May 31, 2025

2025

Sandpaper Socks: Operational Considerations in AI Evals

Olivia Shoemaker

May 31, 2025

2025

Rising Cost of Evaluations, Falling Cost of Intelligence

Ben Cottier

May 31, 2025

2025

Regulation of Frontier Models, Confidential Computing & AI Alignment

Bill Foster

May 31, 2025

2025

Prioritizing Technical Governance Investments: Verification

Robert Trager

May 31, 2025

2025

Policy-Oriented AI Evaluations

Kevin Wei

May 31, 2025

2025

Normal Policy Tools for a Normal Technology

Asad Ramzanali

May 31, 2025

2025

National Security & Defense

Steve Kelly

May 31, 2025

2025

Hydropower: the Missing Piece

Charles Yang

May 31, 2025

2025

Hardware-Enabled Verifiability as a Data Center Security Property

Onni Aarne

May 31, 2025

2025

DoD & AGI Preparedness

Mark Beall

May 31, 2025

2025

Day 2 Opening Remarks

Lennart Heim

May 31, 2025

2025

Compute in America: Playbook for Secure Clusters at Home

Arnab Datta

May 31, 2025

2025

CVE-Bench

Daniel Kang

May 31, 2025

2025

Application of Cryptographic Primitives in AI Accelerators

Fatemeh Ganji

May 31, 2025

2025

Access Considerations for Generative AI Systems Beyond Release

Irene Solaiman

May 31, 2025

2025

AI-Military Integration Under President Trump

Hamza Chaudhry

May 31, 2025

2025

AI Supply Chains: An Emerging Ecosystem of AI Dependencies

Sarah Cen

May 31, 2025

2025

AI National Security Policy: Industry and Government

Sara McNaughton

May 31, 2025

2025

14 Ways of Looking at AI

Brad Carson

May 31, 2025

2025

Unresolved Debates on the Future of AI

Helen Toner

May 30, 2025

2025

Security: Datacenter & Model Weight

Lisa Thiergart

May 30, 2025

2025

Making Sense of the AI Auditing Ecosystem

Miranda Bogen

May 30, 2025

2025

Frontier Model Safety

Rep. Alex Bores

May 30, 2025

2025

Day 1 Opening Remarks

Adam Gleave

May 30, 2025

2025

An Overview of Technical AI Governance

Ben Bucknall

May 30, 2025

2025

AI Control: Addressing Risks from Agentic Internal Deployments

Mary Phuong

May 30, 2025

2025

Media & Attribution

At this event, we used Chatham House Rule for participants and a default off-the-record policy for press, to encourage honest, candid exchange — while selectively greenlighting certain talks for broader distribution to advance the conversation responsibly. Our policies are as follows:

Confidentiality (Participants) & Off-The-Record (Press) Defaults: The event follows Chatham House Rule: participants may use the information received, but neither the identity nor the affiliation of the individuals providing the information or other participants may be revealed. For press, this event is off-the-record, which includes all interactions at the venue as well as all communication taking place on event platforms.

Exceptions:

  • Plenary Room: Speakers in this room may choose to make their remarks on the record, with recordings intended for public release. Whether a session is on- or off-the-record will be clearly indicated to attendees in the room.
  • Press Room: Conversations here are on-the-record by default, unless a participant explicitly requests otherwise.
  • For participants, Chatham House Rule may be waived if all participants in a conversation explicitly consent to being quoted or identified.‍
  • For press, the off-the-record default may not be waived outside of the above two exceptions; participants may move to the press room to make on-the-record comments.‍

Additional Guidelines:

  • Attendee Privacy: The identity of speakers is public by default; the identity of attendees is private by default. The attendee list may not be shared. Journalists may not report attendee identities (other than speakers) without explicit permission.
  • Recording and Photos: Attendees and press may not record or take photos during the event unless the session is on-the-record or all participants have explicitly consented. Photos of attendees may not be published without their consent. Photos of speakers, when taken during their on-the-record remarks, may be shared publicly by default.
  • Press Identification: Press members must wear their press identification badges visibly throughout the event. Press members must maintain clear press identification on all event platforms; organizers will configure press profiles with identifiers that must not be altered.
  • Press Movement: Press may circulate freely throughout the venue and communicate on event platforms, subject to the above policies.
  • Press Designation and Privileges: Only individuals officially registered as press or event organizers may conduct interviews for publication and write articles about the event for news publications or professional media organizations.‍
  • Conduct: All participants are expected to engage respectfully and uphold a professional standard of conduct throughout the event.
  • ‍Violation of the above policies may result in removal from the event, restrictions on future participation, or other actions deemed necessary by the organizers.