Chronos uses a large language model, accessed through our infrastructure partner's AI gateway, to help summarize and organize case documents. Document analysis works on text extracted from the pages you upload. The language model receives that text, not the original file binary; file processing needed to extract the text is a separate step.
AI-generated summaries link back to the source document through citations. Source-grounded Suite outputs include per-line anchors; other outputs use source-page citations. These links let a reviewer compare the generated statement with the record rather than rely on the summary alone.
The in-app interface identifies AI-generated output with draft and review disclaimers so users can distinguish model-generated text from the source record. AI-generated content requires human review before use; a citation does not replace that review.
Chronos uses the model on a prompt/response basis only. Customer data is not used to train or fine-tune any model, by contract and by technical design. There is no Chronos fine-tuning pipeline.
Our Data Practices section describes the related data-handling safeguards.
The system uses grounding checks and OCR-quality indicators to help reviewers assess whether extracted text and generated statements are tied to the source. For Suite claims that go through verification, results are labeled supported, partially supported, or unsupported by the supplied source passages. Unsupported claims are dropped; partially supported claims are marked for review.
These results are NOT a calibrated statistical accuracy probability. They are grounding signals, not a guarantee of correctness, completeness, or clinical accuracy. OCR-quality indicators describe text recognition, not the truth of a statement in the record.
A verification check may use the same model as the original generation. Such results are labeled as same-model checks, not independent verification. Review the cited source even when a statement is marked supported.
The underlying model is selected and configured through our infrastructure partner under HIPAA Business Associate Agreement terms. Chronos also controls application prompts and model settings within the gateway's supported configuration.
We benchmark model behavior against real-world accuracy criteria before adopting prompt or model changes within our control. This process does not establish an immutable version pin for the underlying model, and an automatic change-notification process for infrastructure-partner model updates does not exist today.
Our security overview provides the broader controls and operational context.
Chronos helps organize and review records; it does not provide its own medical opinion or replace professional judgment.