Training Data Summary
This page describes the data Manti collects that may be used to train AI models, and the third-party models the Manti platform currently relies on. We publish it as a matter of transparency and to satisfy the California AI Training Data Transparency Act (AB 2013). We will update this page if our practices change, and before any model trained on the data described below is made available.
Current status: Manti does not currently train AI models
The Manti platform currently runs on third-party foundation models provided by Anthropic, Google, and xAI. Those models are developed and trained by their providers, and information about their training data is available from each provider. Manti does not currently train or fine-tune its own AI models, and no data described on this page has been used for training as of the date above.
Under our Terms of Service, content from Free trial accounts and Plus, Pro, and Max subscription accounts may be used to train and improve AI models in the future. We are collecting the data described below with that possibility in mind. If we begin training or fine-tuning a model, we will update this page before that model is made available.
Data Manti collects that may be used for future training
1. Customer content from Free trial accounts and Plus, Pro, and Max subscription accounts. Chat content, prompts, and project workspace data (including project files, agent session records, error logs and telemetry, and engineering artifacts) submitted by users during a Free trial or on the Plus, Pro, and Max tiers, collected under the license granted in our Terms of Service. This may include content created by users and content generated by AI. Account personal information such as names and email addresses is excluded, and credentials and secrets are removed before collection. Nothing is collected from Enterprise tier or Charter Program accounts. Free accounts are collected from only while a Free trial is active. Collection began in August 2026 and is ongoing.
2. Electronic component reference knowledge. Reference information about electronic components (such as pinouts, footprints, and simulation models) compiled from publicly available manufacturer datasheets and reference sources, built and validated by Manti's systems, and stored in a shared component database. This dataset contains technical reference knowledge only and does not contain customer designs. Compilation began in August 2026 and is ongoing. This data was not purchased or licensed from any vendor.
3. Publicly available open datasets. Openly published datasets, such as those hosted on public repositories including Hugging Face, which may be used for testing, evaluation, and model development under the license terms of their respective publishers.
4. Synthetic data generated by Manti's systems. Manti's systems generate synthetic data during agent operation, including simulation outputs and renders (such as 2D and 3D video and frame images) that are validated automatically by executed code. This data is generated on users' devices during platform use and collected from Free trial, Plus, Pro, and Max accounts.
Additional disclosures
Intellectual property: Customer content may include material in which customers or third parties hold intellectual property rights; it is collected and would be used under the license granted in our Terms of Service. Component reference knowledge is drawn from publicly available manufacturer documentation. Open datasets are used under their respective published licenses.
Purchased or licensed data: Manti has not purchased or licensed training data from any data vendor.
Personal information: Account personal information (such as names and email addresses) is excluded from the data collected for possible training, and credentials and secrets are removed before collection. Customer content may incidentally contain personal information that users include in their own content.
Cleaning and processing: Before any use in training, data would be processed, including anonymization, removal of credentials and secrets, removal of account details, exclusion of account personal information, quality filtering, and validation.
Synthetic data: Yes, as described in item 4 above.
Time period of collection: Customer content, component reference knowledge, and synthetic data have been collected from August 2026 onward on an ongoing basis. Open datasets are as published by their sources.
First use in training: None as of the Last Updated date. Manti has not trained or fine-tuned any model using the data described on this page. This page will be updated before any such model is made available.
Questions about this page may be directed to privacy@mantiai.com.