Project workspaces use AI to identify the claims in the deliverables you upload. This notice explains what is sent to our AI provider, how it is handled and what you should check before you upload.
Which features use AI
- Claim extraction in project workspaces. When you upload a primary deliverable to a project, AI can identify the claims it contains so you can review them against your references.
- Quick Annotation does not use AI. PDF viewing, annotation, labels, folders and sharing run entirely on our own systems.
- Evidence matching and numerical checks, where available, use rule-based methods on our own servers and do not send content to an AI provider.
What is sent and when
- Your uploaded deliverable is converted to text on our servers.
- When you process a project, the text passages of the deliverable are sent over an encrypted connection to the OpenAI API, together with instructions describing what to identify.
- The claims returned are stored in your project, where you can review, filter and check them against your sources.
We send document text and processing instructions only. We do not send your name, email address, payment details or other account information. Supporting references are not sent to the AI provider.
Our AI provider and how it handles data
- Our AI provider is OpenAI, L.L.C., used through its business API.
- Our requests ask OpenAI not to store responses for later retrieval.
- OpenAI does not use API data to train or improve its models.
- OpenAI may keep inputs and outputs in abuse-monitoring logs for up to 30 days, unless the law requires longer, and may briefly cache encrypted processing data for up to 24 hours to speed up repeated requests.
- OpenAI processes data in the United States and other countries where it or its sub-processors operate.
Accuracy and human review
AI output can miss claims, include text that is not a claim, or describe a passage incorrectly. Treat every result as a starting point for review. AI output is not medical, scientific, legal or regulatory advice, and it is not an approval of any claim. Always confirm claims against the original sources before relying on them.
Credits and estimates
Where paid plans are offered, AI processing uses AI Processing Credits. Each project shows an estimate, and credits are reserved only after you approve it.
Your choices
- AI claim extraction is part of project processing. To keep a document away from AI processing, annotate it in Quick Annotation instead of uploading it to a project.
- You can cancel processing that is still running from the project overview.
- Text that has already been sent cannot be recalled, and claims already created stay in your project until you delete the project or your account.
Client and confidential materials
Before uploading documents to a project, check that your agreements with your clients or employer allow their documents to be processed by third-party service providers such as OpenAI. Do not upload documents containing identifiable patient health information.
Questions
Contact support@easyannotation.com with any questions about AI processing.