The policy below is intended to be used within cause-based organizations which wish to encourage the experimental use of ChatGPT or other LLMs within their organizations. It is intended to be repurposed, expanded, and otherwise modified as needed. Many thoughtful individuals interested in the use of LLMs for politics contributed to this text, but improvements are of course always possible.
Other resources related to this one are noted below, you may want to consult these as well.
The sample policy is released under the MIT License as noted below.
It’s clear that ChatGPT and other Large Language Models (LLMs) like Bard may well have a tremendous impact on our future. We encourage you to experiment with these tools! However, please stick to some key guidelines to avoid any potential harm to our organization or the people we serve:
- We wish to encourage your use of LLMs for experimental purposes, and to discover novel innovations which may propel our organization forward. This policy is meant to promote that kind of usage, and we welcome any feedback about it. In particular, we welcome feedback about the need for “ChatGPT Plus” licenses or similar resources if the reason is compelling.
- Do not disclose sensitive proprietary material, such as strategic memos, lists of donors, or sensitive constituent data, to third-party services like ChatGPT or Bard. It is crucial to remember that we’re responsible for protecting the data we manage. In cases where LLM use might involve data that could be sensitive or personally identifiable, stakeholders should be informed of potential risks.
- When using the by-products of an LLM in your work, please properly cite it. (For example, “when I asked ChatGPT what the weather will be tomorrow, I got: …”.)
- When using the by-products of an LLM in your work, you are ultimately responsible for those by-products. You should therefore verify any factual claims, check for logical and grammatical errors, and ensure that the content adheres to our organization’s standards and values.
- Be cognizant that there are ethical concerns with the use of large LLMs for a variety of reasons. For example, ChatGPT has been criticized for being trained on racially-biased text, and fed data provided by exploitative labor practices. Think carefully about how such criticisms might reflect on our work.
- Do not use ChatGPT or Bard for production-quality systems, especially at the free tier. We consider these systems to be potentially useful augmentations to our internal work rather than a mission-critical and resilient system.
- Take care in adhering to the ChatGPT or Bard usage policies. In particular, note that large-scale use of ChatGPT for political purposes is forbidden. Unfortunately, the full meaning if this policy has not yet been made clear.
- If you intend to use LLMs on an ongoing basis in some clearly-defined process, or if you intend to develop a program around LLMs, please report it to your manager.
- Please share this policy with any vendors who may have access to sensitive orgnizational materials, and ensure that they do not share these materials with cloud-based LLMs.
- This policy may be updated from time to time, both in response to new developments in LLM technology and as we learn from our experiences. We will make it a priority to keep you informed of these changes and invite open discussions around them.
- We view AI tools like LLMs as resources meant to augment and enhance the work of our professionals, not to replace them. It’s essential to remember that human skills, intuition, and judgement are invaluable. As such, while we encourage the innovative use of AI, we also urge the continued development of human-centric skills.
Sample AI transparency label
“The content of this email was generated with the help of an AI Large Language Model (LLM), and carefully reviewed by our team for accuracy and alignment with our campaign values.”
- Five steps executives should take for the ethical use of AI
- How to create the best ChatGPT policies
- ChatGPT Policy Template
- Google’s White Paper on AI Policy
Copyright 2023 Shai Sachs
Permission is hereby granted, free of charge, to any person obtaining a copy of this text (the “Text”), to deal in the Text without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Text, and to permit persons to whom the Text is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Text.
THE TEXT IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE TEXT OR THE USE OR OTHER DEALINGS IN THE TEXT.