Partnership on AI: The First Ten Years
PAI founded by five major tech companies—Amazon, Facebook, Google, IBM, and Microsoft—to ensure AI development advances positive outcomes for people and society
The New York Times covers PAI’s founding
“Protecting Humans and Jobs from Robots Is 5 Tech Giants’ Goal”
“[PAI] is a huge step forward, breaking down barriers for AI teams to share best practices, research ways to maximise societal benefits and tackle ethical concerns.”
Mustafa Suleyman & Greg Corrado, Google
First All Partners Meeting in Berlin establishes PAI’s core values to guide strategies and actions
First working groups establish priority research topics
- Fairness, Transparency, and Accountability
- Safety-Critical AI
- AI, Labor & the Economy
- Social & Societal
Second All Partners Meeting
Partners met in San Francisco to discuss PAI’s progress and future goals
Human-AI Collaboration Framework and Case Studies published
AI & Media Integrity Steering Committee formed
With the BBC and WITNESS, PAI’s media integrity work begins with the Deepfake Detection Challenge
C2PA is envisioned at PAI’s third All Partner Meeting
Conversations between PAI members inspire the creation of the Coalition for Content Provenance and Authenticity (C2PA)
AI Incident Database launched
With more than 1,000 reports of AI failures, the AIID is an essential tool for AI researchers and developers
PAI takes a stand against AI misuse
In response to a racially biased risk assessment tool, PAI published “Algorithmic Risk Assessment and COVID-19: Why PATTERN Should Not Be Used”
Manipulated Media Labeling Principles published
12 principles designers should follow when labeling manipulated media, published in collaboration with First Draft
The National Security Council cites ABOUT ML in their recommendations to Congress, highlighting it as a model for responsible AI documentation.
Four major resources published
- Responsible Sourcing of Data Enrichment Services
- Publication Norms for Responsible AI
- ABOUT ML Reference Document & Resource Library
- Fairer Algorithmic Decision-Making and Its Consequences
Global Task Force for Inclusive AI announced at the White House Summit for Democracy
Rebecca Finlay hired as CEO
New Five-Year Strategic Plan
Dozens of partners engaged in developing and adopting PAI’s process
Google DeepMind adopts PAI’s data enrichment sourcing recommendations
Meta cites PAI’s publication norms for OPT-175B release
Three ABOUT ML pilots examine the AI transparency practices of UN OCHA, Intuit, and Biologit
The White House’s AI Bill of Rights cites PAI’s ABOUT ML work
Synthetic Media Framework launches
With 10 founding supporters, prescient timing just as ChatGPT hits mainstream
Guidelines for AI and Shared Prosperity published
Guidance for Safe Foundation Model Deployment released
PAI enters the policy sphere
Partners gather in New York City at the first PAI Policy Forum, fostering formation of the Policy Steering Committee. PAI then brings the discussion to Bletchley Park.
Synthetic Media case studies published
16 organizations reported on their adoption of PAI’s Framework, including Adobe, OpenAI, the BBC, Meta, Microsoft, and TikTok
PAI’s policy work garners multiple citations from NIST, the US Department of Labor, USAID, the US Department of State, and NTIA
Beyond the Code video series produced in collaboration with BBC StoryWorks
PAI participates in several international AI events
- Europe: AI Action Summit
- Asia: Asia Tech x Singapore
- Africa: Deep Learning Indaba, Global AI Summit on Africa
European Steering Committee launches
PAI’s first international steering committee
PAI prepares for the rise of agentic AI
- Prioritizing Real-Time Failure Detection in AI Agents
- AI Agents & Global Governance
- Preparing for AI Agent Governance
PAI contributes to drafting the EU AI Act Code of Practice
New Strategic Plan adopted, positioning PAI for an influential future
Partner Forum in Geneva during the UN’s AI for Good Global Summit
AI and Human Connection workshop strengthens PAI’s focus on people and society
Celebrating ten years
PAI enters its second decade of impact