0. General Information

Note

In the first stage of the application form, you will be asked to describe your digital solution. Please ensure that the details you provide, links to your website, source code, and all other documentation are available in English, so our technical review team can assess the application properly. Thank you.

---

"The UN SDGs are a collection of 17 interlinked global goals designed as a blueprint to achieve a better and more sustainable future for all, addressing the global challenges we face, including poverty, inequality, climate change, environmental degradation, peace and justice." An eligible digital public good must be designed to solve a shared global common problem for one or more groups of people with direct relevance to one or more SDG targets. The matrix chart below illustrates how we evaluate potential DPGs for SDG relevance, irrespective of their industry:

Typically, only solutions in between the 1st, 2nd, and 4th quadrants are eligible (nothing that some will fall in between). Solutions like generic libraries, plugins, frameworks, programming languages, interpreters, compilers, firmware, protocols, etc., that are explicitly targeted to software engineers would have a low direct SDG impact and low breadth and hence are not eligible to be certified as a digital public good.

---

Jump to DPG Category ↓

• 🖥️ Open Software
• 📊 Open Data
• 🤖 Open AI System
• 📚 Open Content

---

Open Software

We accept most types of source code for software solutions built using different kinds of technology.

✅ Examples of Acceptable Solutions	❌ Examples of Ineligible Solutions
Web applications. Mobile applications. Desktop software applications. Operating systems.	Generic libraries, plugins, packages, and frameworks. Firmware and device drivers. Programming languages, interpreters, compilers, etc. Protocols.

---

Open Data

We accept one-off releases of a single dataset, a collection of datasets, or an API.

✅ Examples of Acceptable Solutions	❌ Examples of Ineligible Solutions
Historical or statistical data that can be accessed or downloaded directly through a URL. Topic or sector-related data aggregation platforms or collections, as long as all data is open. A public API operating over static or real-time data. Other open datasets like geographical, public sector, regional, science, education, environmental, or other types of data.	Graphics or visualizations (static or interactive) without access to the raw data. High-level aggregated data or single data points. Any datasets not accessible in open formats.

---

Open AI System

We only accept AI Systems that contain the following components:

1. Dataset(s) used to train the system.
2. Code Components:
   • Data pre-processing.
   • Training, validation, and testing.
   • Inference.
   • Supporting libraries and tools.
3. Model Components:
   • Model architecture (the type of model, layers, and structure).
   • Model parameters (weights, optimizers, coefficients, and other applicable hyperparameters).

Note

These components are based on the Model Openness Framework (MOF) and Open Source AI Definition Checklist. Other components listed in these frameworks, such as research papers, evaluation results, and sample model outputs, among others, are optional unless specified in other indicators of the DPG Standard.

Important

Software solutions that use specific third-party AI components like LLMs to deliver specific product features are NOT considered AI Systems but still Open Software. Consider the following questions:

• Does the system include a model that was trained explicitly on data (or finetuned with data) to infer behaviour (developers define how the system learn rather than what it does) rather than explicitly programmed (developers write step-by-step instructions)?
• Does the system include a model that adapts its outputs or behaviour over time without reprogramming?

If your answer is YES to questions one and/or two, then your solution is an AI System.

✅ Examples of Acceptable Solutions:

❌ Examples of Ineligible Solutions

Fully open, task-specific (narrow) AI systems where the training datasets, code components, and model components are openly licensed, available, and properly documented (including any supporting software/hardware user interface for the model usage). Fully open generative AI systems, where both the initial training datasets and any additional finetune or continual learning datasets, code components, and model components are openly licensed, available, and properly documented (including any supporting software/hardware user interface for the model usage).

Academic paper (theoretical). Software where the usage of the AI model is not the main component (this will fit under the regular software category).

---

Open Content

We only accept open content collections and do NOT accept individual pieces of content.

✅ Examples of Acceptable Solutions	❌ Examples of Ineligible Solutions
Collections of learning materials or resources. Collections of publications or academic papers. E-learning open content that addresses various topics. Collection of books, guides, manuals, etc. Other collections of content like data visualizations/aggregators, blog posts, infographics, audiovisuals, images, audio, and other forms of digital content.	Individual pieces of content. Social media posts. Single book, guide, article, post, media, etc.

---

Tip

Here's a collection of extra resources and helpful links curated by the DPGA and the DPG community you can explore or contribute to.