Why we have this policy
This policy sets out how we use AI in service of our mission. We use AI to support the creation of a more just and healthy society including solving problems, supporting the people dedicated to this work, and amplifying shared AI learning to accelerate our collective vitality and impact.
Who it covers
Staff and contractors when they're working for us, including when they use their own AI tools on our behalf.
Terms we use
- AI (artificial intelligence) is technology that learns patterns from large amounts of data and uses them to recognize, predict, create, or act.
- Assistive AI suggests or predicts while you work.
- Generative AI creates new content from your instructions (prompts).
- Agentic AI carries out tasks on your behalf, often several steps at a time.
- Built-in AI is AI functionality inside tools we already use, such as graphic design, search, and database software.
- Standalone AI applications are dedicated AI tools, including large language models such as ChatGPT, Claude and Gemini.
- Personal information is anything that identifies a real person or tells you something private about them.
Our principles
- 1. AI as a mission multiplier. We use AI to increase sustainable capacity-building and support staff in mission-driven organizations. We apply AI to measurable challenges and to support strategic priorities, not to automate ineffective processes or inequitable practices.
- 2. Maintain discernment. Everyone is responsible for verifying their output including AI- generated or -assisted products and decisions. This includes ownership of quality, truthfulness, and a basic understanding and countering of bias present in AI systems.
- 3. Uplift other people. We collaborate with those doing the work. Improving the quality of work-life for staff is as important as improving output; both are valuable outcomes required for sustainability. Everyone can learn and upskilling is a goal on par with implementation.
- 4. Be transparent. We're open about how we use AI, including by publishing this policy. We proactively share how we protect your information and work to maintain trust.
- 5. Advocate for responsible AI. We are active participants in shaping the use of this transformational technology. We continually seek to understand this rapidly changing field, speak plainly about risks, and advocate for responsible use and controls.
Approved tools and accounts
Standalone LLM tools used on a daily basis
| Tool | Approved for uses | Account type | Training on our data turned off? |
|---|---|---|---|
| Claude | Generative, and Agentic AI | Organization | Yes |
| Gemini | Assistive, Agentic AI in Google apps, Image Generation | Organization | Yes |
| Lovable | Website Design, Development and Hosting | Organization | Yes |
AI functions built into approved tools
We use AI functions built into paid applications within our organization or at client sites on a regular basis per each project’s requirements. Examples include and are not limited to: Canva, Hubspot, Microsoft Office, Salesforce.
Other AI tools
Different AI tools, via free and paid accounts, are used for experimentation using public content. Whenever possible, we opt out of training data and use two-factor authentication.
Account security best practices
- Use organization accounts for work, not personal ones.
- Turn on two-step login
- Don't share passwords.
- Only those requiring access receive access for the duration of the need; clients explicitly approve access by any member of our company or our contractors.
Personal and data privacy
- We do not enter private personal information about donors, volunteers, clients, or staff into standalone AI applications.
- We do not enter confidential organization business, such as unreleased financials, legal issues, or personnel decisions into standalone AI applications.
- We only use AI within paid subscriptions that enable us to turn off use of our data for training.
- We do not generate AI images of real people; any alterations must preserve the original intent and context of the image.
This policy adds to, and never replaces, our existing privacy and confidentiality policies, our contracts and grant terms, the professional ethics codes we follow, and the laws that apply to our work. Where any of these is stricter than this policy, the stricter rule wins.
Transparency and disclosures
- We disclose when AI produces or significantly contributes to content, by providing attribution on the relevant asset.
- We disclose and ask for permission when AI takes part in an interaction, such as a chatbot or an AI note-taker.
If something goes wrong
If personal information goes into an AI tool accidentally, an AI error gets published, or an output seems biased, tell the Owner right away. Reporting quickly is what matters and we will work to transparently address the error.