Candid artificial intelligence notice

At Candid, we strive to deliver valuable products, services, and content to our users, and may utilize AI tools to supplement the user experience. We are also committed to transparency and responsible artificial intelligence (AI) usage. Below is a summary of the AI solutions we use to enhance our products and services, the steps we take to manage potential risks, and the data we collect and process as a result. 

AI solutions and purposes 

We currently deploy the following types of AI solutions to provide, enhance, and/or customize our products and services: 

• Candid’s AI Assistant: Get quick answers to your fundraising and nonprofit sector questions by submitting queries to our AI Chatbot. Our AI assistant (currently in beta) will use curated resources to answer questions and offer tailored recommendations. 

• RFP Recommendations: Boost your chances of securing grants with tailored recommendations on open requests for proposals (RFP) opportunities. Login to our platform to identify open funding opportunities relevant to your work based on information you share with us. 

• Funder Recommendations: Expand your network and let our tool help you find new funders for your nonprofit. Simply login to our platform and provide information like your organization name, EIN, program description and location to identify prospective funders.  

• AI-Powered LOI Writer: Craft compelling letters of intent (LOIs) with ease by using our AI-powered to tackle common pain points in grant writing based on best practices, including factors like grammar, vocabulary, and sentence structure. 

• Social Sector News Feed and Philanthropy Classification System (PCS): Candid uses machine learning to automatically tag new grants, RFP announcements, award announcements, press releases and news articles to our PCS system to offer a detailed view of activity within the sector and enable our stakeholders to accomplish their goals (e.g., seek funding). 

The majority of the above AI solutions are available via Candid Labs, through which we design and test our latest innovation projects. Users are invited to participate in beta testing to provide feedback on their experiences so we can further improve the tools and develop meaningful solutions for our stakeholders. Please visit this page for additional information on Candid Labs and its AI-based innovation solutions: https://candid.org/labs.  

AI development and training

Candid has engaged a team of experts to internally develop its AI solutions. Enclosed below is a high-level summary of the datasets used in the development of our generative and non-generative AI systems and solutions.  

• Dataset Sources: Our AI systems and solutions were developed using a mix of publicly available datasets, open-source datasets, licensed proprietary data (such as the news we share in our social sector news feed as described above), customer registration data, user-contributed data, and other data sources. The datasets do not contain unlicensed materials protected by copyright, trademark, patent or other intellectual property protections. 

• Data Acquisition and Collection Timeline: Where necessary, datasets were (1) licensed or purchased under commercial agreements, (2) obtained through open-source access or public domain contributions, (3) contributed by our users, or (4) collected from customers at registration, per our Privacy Notice and Terms of Service. Data collection is ongoing for quality assurance and AI improvements. 

• Data Types: Our AI system and solutions consider the following types of data to generate its outputs: (1) usage information, such as the user prompts and other inputs entered into the system (e.g., request for proposal (RFP) responses); (2) organizational and profile information, such as the data the user and/or organization enters into the system regarding its mission and organizational structure, budget, and other details; (3) public information, such as news and blog posts relevant to our business and customers; and (4) financial information, such as financial data received via Form 990. 

• Personal Information, Aggregate Consumer Information, and Synthetic Data: Given the nature of the AI systems and solutions Candid is now offering and/or embedding in its products and services (as described above), we designed the AI systems and solutions to consider the privacy personal information and/or aggregated consumer information where appropriate (e.g., per our Privacy Notice). For the moment, Candid has not relied on synthetic consumer generation to simulate use cases, improve resilience to edge cases and adversarial prompts, and otherwise test and evaluate accuracy and quality assurance.  

• Purpose Alignment: These datasets were selected to support the ability of our AI systems and solutions to respond to diverse user inputs and generate coherent, contextually appropriate material. The data points involved reflect those Candid employees would need to consider to generate similar content upon request. For example, using an organization’s mission, profile, financial information, and current RFP search queries and/or responses to recommend similar RFPs to apply to for funding. 

• Dataset Processing: Datasets undergo data cleaning, normalization, and deduplication, with the goal of enhancing accuracy, quality assurance, and AI safety and utility. 

Overview of non-consequential AI use  

Candid does not currently use AI to automate, accelerate, or inform consequential decisions. None of Candid’s AI systems or solutions generate or drive decisions with material legal or similarly significant effect on its customers or other individuals. Indeed, Candid’s AI solutions focus on generating informational responses, recommendations, and/or draft materials to inform user actions and downstream decision-making rather than engaging in any direct decision-making, internally or externally. Individual users are encouraged to critically review any draft materials or recommendations, per our terms and conditions, before utilizing the material.  

Similarly, none of the AI-enabled services offered by Candid intersect with the subject-matter areas commonly considered high-risk, such as education, employment, financial or lending services, health-care services, housing or insurance. Nor does Candid use AI to evaluate, analyze, or predict personal aspects related to an individual’s economic situation, health, demographic characteristics, personal preferences, interests, reliability, behavior, location, or movements (i.e., “profile” individual consumers).  

Candid therefore does not offer mechanisms to opt out of such processing or appeal automated decisions. 

AI governance and risk management

As part of its commitment to responsible AI usage, Candid has implemented an AI governance and risk management process that includes risk management techniques like (1) risk assessments (e.g., bias, algorithmic discrimination, and fairness testing) and (2) quality assurance measures, such as human oversight of critical AI functionality and outputs.  

Contact us

If you have any questions about our practices or this Notice, please contact us at contracts@candid.org.