| CSISCtr. for Strategic & International Studies · Washington, DC |
- LLM benchmarking (Futures Lab) — Pentagon CDAO-funded; tests GPT-4o, Claude, Gemini, DeepSeek, Llama, Qwen on geopolitical decision-making; AI and human experts produce parallel forecasts
- CFPD-Benchmark — CSIS's proprietary tool for evaluating LLM bias in foreign policy decisions; open for other researchers to use
General staff AI productivity tool use not publicly documented. Prior claim about routine internal use of ChatGPT/Claude/Gemini was unverified and removed.
|
— Not found
No published internal staff AI use policy. Wadhwani Center produces AI governance guidance for governments — not an internal staff policy.
|
Leading
Wadhwani AI Center is among DC's most influential AI policy authorities; provides Congressional testimony, advises US government, publishes annual AI Policy Forecast. CFPD-Benchmark is a publicly available AI evaluation tool.
|
CSIS LLM benchmarking study ↗
NPR: CSIS AI diplomacy ↗
Wadhwani Center ↗
|
| Atlantic CouncilWashington, DC |
- Microsoft AI tools — Azure AI via Global Stage co-production partnership; specific tools not publicly named
- DFRLab open-source AI tools — custom and open-source models for detecting AI-generated disinformation, deepfakes, and synthetic media campaigns
General staff AI productivity tools not publicly documented.
|
— Not found
No published internal AI use policy. Annual report references AI governance research outputs, not staff rules.
|
Very Active
GeoTech Commission on AI (Nov 2025) sets the US AI agenda across six domains. AI Connect program funds AI governance capacity in developing nations (State Dept. grant). DFRLab is a global authority on AI disinformation and synthetic media.
|
GeoTech AI Commission ↗
AI Connect program ↗
2024 Annual Report ↗
|
| Brookings InstitutionWashington, DC |
- General LLMs (ChatGPT etc.) — 40% of staff were already using AI tools at time of policy publication (Jan 2024); permitted uses include brainstorming, editing, summarising, drafting social posts, generating images
No specific tools mandated or approved by name in published policy. General-purpose LLMs permitted with restrictions on sensitive data and PII.
|
✓ Published
"Provisional Principles for the Use of Generative AI" (Jan 2024) — publicly available on brookings.edu. Covers all research and operational work. Developed by an all-staff ETAG advisory group. Explicitly bans inputting sensitive data or PII. Published to serve as a template for peer organisations.
|
Active
AIET initiative produces AI governance recommendations for policymakers. Global Task Force on AI in Education (2024–26). Regularly cited in Congressional AI debates and global governance forums.
|
Brookings AI policy (Jan 2024) ↗
ETAG approach article ↗
AIET Initiative ↗
|
| CFRCouncil on Foreign Relations · New York |
- LEAD AI programme tools — new CFR programme (2025–26) uses AI-assisted research for national security and democratic resilience analysis; specific tools not named publicly
No specific staff AI tools publicly documented.
|
— Not found
No published internal AI use policy found for CFR staff.
|
Active
LEAD AI programme (launched 2025) focuses on democratic AI governance across four pillars. Foreign Affairs publishes major AI strategy pieces. Digital and Cyberspace Policy Programme is influential in DC AI debates.
|
CFR LEAD AI programme ↗
CFR: Regulating AI ↗
|