Partnership on AI (PAI)
- Type
- Civil society organisation
- Place
- International — International (US-based non-profit)
- Last checked
- 2026-09-22
- Next check due
- 2027-03-21
Ways to reach them
- Contact form
- Become a partner organisation
This route is retained as reference only. Its verification is missing, more than 180 days old, or the contact window is not currently open; do not rely on it until it is re-checked.
- Homepage
What it does
Multistakeholder convener with 100+ partner organisations spanning Amazon, Google, Meta, Microsoft, OpenAI, Apple alongside ACLU, ARTICLE 19, Future of Privacy Forum, UNICEF, UNDP and universities. Produces guidance documents and frameworks — notably on safe foundation model deployment, synthetic media, and AI and labour — across four programme areas: AI and Human Connection; AI, Labor and the Economy; Safety-Critical AI; and Public Policy.
Honest assessment
Consensus-bound. With the largest labs as partners, PAI's outputs are negotiated to the level everyone can accept, which limits how sharp a concern it will carry. Organisational partnership is possible; individual input is not really a thing. Middling responsiveness.
Concerns it covers
Deliberately plural. Its Safety-Critical AI programme touches frontier deployment risk; its labour and human-connection work is squarely near-term/structural. Its distinctive contribution is process norms rather than a risk thesis.
Government channel
Moderate — PAI guidance is cited in policy processes and it participates in multistakeholder forums, but it is a norm-setter, not an advocate.