AI Alliance

Introduction

The mission of the AI Alliance is to accelerate and disseminate open innovation across the AI technology landscape to improve access and participation in the development of foundational tools, capabilities, safety, security, and trust in AI, and also to responsibly maximize the benefits of AI for people and society everywhere. The Alliance was founded on the conviction that an open and responsible approach to AI technology, based on the principles set forth below, will provide immense benefits to society.

Advances in AI over the last decade, powered by availability of training data, vast computing infrastructure and innovation in machine learning algorithms, have made it possible to perform a variety of tasks; including code, text, image, audio, and video understanding and generation, along with digital web agents; enhancing human potential, and increasingly exceeding human performance, leading to major advances in medicine, education, disaster mitigation, etc. The rise of generative AI using foundation models (e.g. large language models) has sparked widespread excitement for its promise to solve a wide range of business and societal problems. However, despite this opportunity, we risk limiting the widespread realization of this technology's benefits if we do not actively encourage and enable open innovation. Furthermore, current architectures for generative AI will someday be replaced by new, more capable architectures that will emerge in the coming years, and our aim is to give equal access and opportunity to empower the creator and user communities around the world to participate in this journey.

Openness and Access

Open innovation in AI should be encouraged and should not have disproportionate regulatory burdens placed on it.
Researchers, developers, and deployers should be encouraged to innovate, develop, and have access to AI artifacts that help build and run models more efficiently, fix bugs, implement mitigations, and make the whole ecosystem more sustainable and resilient.

Selection and Choice

A broad selection of AI models, datasets, and tools, should be made available to empower developers, deployers, and users to innovate using the state-of-the-art AI technologies best suited for their use case.
Developers should be encouraged to select the highest level of openness and compliance that meets the needs and context of their projects.

Safety and Security

Safety and security are the responsibilities of everyone in the AI supply chain including model developers, model deployers, and users.
Where possible, developers should enable community scrutiny of AI models, datasets, and benchmarks from diverse stakeholders, and support the community development of tools and capabilities to help enhance safety, security, risk management, and help prevent misuse.
Education and investment into raising awareness and building competence in the development and use of AI safety and security tools is essential to ensure that as AI proliferates, adoption and use does not cause avoidable harms.

Privacy and Transparency

Developers and deployers should prioritize research on the societal risks posed by AI systems, including avoiding harmful bias and discrimination, and protecting privacy and broader human rights.
Transparency throughout the AI research and development lifecycle builds trust and supports more widespread adoption of investments in AI. Degrees of openness will naturally emerge based on what makes sense for the respective contributors to the open source AI ecosystem.

Economy and Development

The AI community should encourage openness in the development of AI to enhance competition, expand user choices, increase product diversity, and stimulate innovation throughout the AI value chain.
Startups play a critical role in creating and strengthening the emerging open innovation AI ecosystem, which in turn supports economic development.

Societal Impact and Diverse Viewpoints

AI development should include diverse perspectives throughout the AI software and hardware development lifecycles, including those of policy makers, regulators, users, and all other stakeholders in the AI ecosystem; open innovation has a unique role to play in enabling this feedback.
The AI community should encourage and support socially beneficial applications of AI.
The AI community should cultivate inclusive approaches to AI development.