AI Features & Use of Peter

Peter is an artificial intelligence system built on Anthropic's Claude models. It is not a human, and its outputs are generated using machine learning, which may change over time as the underlying models improve.

Peter's outputs — including statistical interpretations, suggested conclusions, summaries, and any generated text or visualizations — are provided for informational and productivity purposes only. They may contain errors, omissions, or inaccuracies, and are not a substitute for your own scientific judgment, peer review, or statistical validation.

You are responsible for independently verifying any AI-generated output before relying on it in publications, regulatory submissions, or any decision with safety, financial, or legal consequences.

Peter can take some actions directly within the platform — such as running an analysis, creating a graph, or organizing a dataset — rather than only answering questions.

Actions that modify or delete existing work — updating or deleting a graph, or resizing your plate — require your explicit confirmation in chat before Peter executes them. Other actions, such as creating graphs or slides, adding analysis nodes, or organizing data, may run automatically as part of normal use.

Peter only has access to your own account's data and can only take actions within the permissions granted to your account. We maintain safeguards designed to resist manipulation of Peter through malicious inputs, including data uploaded into the platform, but no safeguard is perfect — you should not upload data you do not want processed by an AI system.

When interacting with Peter, you agree not to:

• Attempt to extract Peter's underlying system instructions, prompts, or configuration.
• Attempt to manipulate Peter — via direct input or content embedded in uploaded data — into taking unauthorized actions or bypassing confirmation safeguards.
• Use Peter to generate or process content that violates Platelet's general terms of use.
• Use Peter in a way that attempts to access, infer, or reconstruct other users' data.

We reserve the right to suspend or restrict AI feature access for accounts that violate these terms.

Peter is powered by Anthropic's Claude API. When you use Peter, your prompts and a summary of your data's structure (such as feature and reading names) are sent to Anthropic to generate a response.

Raw, well-level plate data is not sent to Anthropic — analyses you ask Peter to run are performed locally on our own servers, and only the results are shared back into the conversation. Anthropic's handling of the data it does receive is governed by our agreement with them as a subprocessor.

Peter conversations are not stored. They exist only in memory for the duration of your session and are discarded once the conversation ends or the page is refreshed — we do not keep a database of chat history.

You retain ownership of your underlying data. Outputs generated by Peter based on your data — analyses, summaries, suggested interpretations — are yours to use, subject to the limitations on accuracy described above.