In February 2023, PAI launched the Responsible Practices for Synthetic Media: A Framework for Collective Action, with 18 partners to date aligning their practices for ethical and responsible development, creation, and sharing of AI generated media.

As generative AI technology propelled realistic synthetic media from the lab to the laptops and devices of everyday people, PAI’s Framework was already collaboratively defined by and available to key stakeholders–like builders of infrastructure and technology, creators, and media distributors and publishers–whose decisions influence the societal impacts of these technologies.

As synthetic media becomes one of the engines of digital content creation, the responsible practices advanced by Framework supporters have the power to improve consent, disclosure, and transparency for massive end-user bases. With 18 organizations signed on as institutional supporters of the Framework, we are seeing PAI’s principles of synthetic media governance in practice, scaling our impact to the billions of users reached by Framework supporters such as BBC, Google, Meta, and TikTok.

In 2024, PAI will publish in-depth case studies from Framework supporters on how they have applied Framework principles in practice, and use learnings to inform public policy.

PAI launched Guidance for Safe Foundation Model Deployment, incorporating expertise from 50+ experts and model providers on how to responsibly develop and deploy foundational AI models in ways that promote safety for society.

The foundation models that are powering this wave of innovation in generative AI are a capability step-change from previous technologies. At the cusp of mass adoption, with a growing range of scenarios and unanswered questions around how these models will be monetized and their real world implications, PAI’s guidance fills a critical knowledge gap by articulating the landscape of AI risks and how to address them while pursuing model innovation, encompassing different model types and release strategies.

As discussions on AI safety began to heat up among global policy makers, our Model Deployment Guidance laid the groundwork for convening our first-ever AI Policy Forum and joining time-sensitive policy conversations like the UK AI Safety Summit with global leaders working to build the foundation of good AI governance and safety for society.

In 2024 PAI will refine recommendations for open access models and publish an updated version of the Guidance.

PAI published Guidelines for AI and Shared Prosperity, with 13 endorsements from experts and stakeholders affirming the power of these recommendations to guide AI deployment in ways that benefit all.

2023 propelled us into a world where 79% of global business and technology leaders expect AI to significantly transform their organizations within the next 3 years, and more than 50% of those same leaders also expect generative AI to centralize economic power and increase inequality. But we don’t have to build that future.

PAI’s Guidelines summarize actionable ways AI-creating and AI-using organizations, labor organizers, and policy makers can assess and ground their decisions, agendas, and interactions in a systematic understanding of the potential impact of AI on the labor market and job quality, centering worker voice in AI development and use.

In this moment of profound opportunity to intentionally design AI technology for a more just, equitable, and prosperous world, changes in practice like OpenAI’s commitment to PAI’s Data Enrichment Sourcing Guidelines for development of ChatGPT4 and iterations illuminate a path to shared prosperity. In this case, ensuring data workers experience working conditions that respect and value their contributions.

In 2024 PAI will work closely with labor organizers and industry partners to increase adoption of the guidelines, and to develop standards of practice for fair treatment of data workers.