How is my data used by Toolpath?

Toolpath does NOT use any tool data or part geometry from your files to train our AI models.

Toolpaths AI models are based on a reinforcement learning method that doesn't require your data directly. While some AI's like ChatGPT style models or Tesla's Autopilot rely on huge amounts of data that is effectively crowdsourced, Toolpath's AI approach doesn't require that at all.

We've built a game-based AI similar to something that would play Chess, Checkers, or Go. These models are not trained by looking at a history of every single chess game every played. Instead, they are given a virtual chess board and set to play against another AI opponent again and again, and getting better each time. Our AI works similarly. We've built a virtual board game to represent machining a part, and built an AI that is good at playing that game. Each of your parts become a unique instance of that game that gets played, but don't get sucked back in to retrain the AI.

So your data remains your data. We don't use it to train Toolpath. If you upload your own library of tools, they won't be used to help make another shop good at using those tools. We're not collecting data on your feeds/speeds or your cut-configurations. We're not continuously re-training our machining game AI on your parts.

There are only two notable exceptions: 

1) When you specifically provide us with data to report a bug, and give us permission to look at it. In that case, we will look at that specific part/tool and develop our own data derived from yours for that specific issue. This is only done with your approval.

2) If we encounter a hard-error while processing your part, we have the ability to replay that part and tool library through the system to replicate the error and investigate it. When we do this, we anonymize the data so the development team doesn't know which customer it came from.

If you're not using our data to train the AI, how does it get better over time? 

We have multiple ways to judge the quality of the estimates and machining programs that Toolpath created. We have our own R&D shop in Bedford OH where we actively test our work on our own machines, but we also have an active community of users who are not shy about telling us when they think Toolpath got it wrong. 

When users share their feedback with us, (sometimes they do so with actual part files and tool libraries, and sometimes more abstractly) we develop new cases to challenge the AI with. But this is done with your explicit permission. So anything you don't want to share, you don't have to. 

Toolpath's data retention policy is detailed below:

• When a part is deleted from your Team from the web interface, the data file and tool library data are anonymized and retained for potential use in debugging or quality assurance. This is not used for training the AI, but only to be able to replicate bugs or errors and correct them. 
• Toolpath can manually delete a user account if requested, which will wipe all of a users account data (name, email, password, etc) - but not any parts, tool libraries, or cut config data unless specifically requested. All part, library, and cut config data upload by a user belongs to the Team and are retained by default, even when that user is removed. 
• If a Team administrator requests it, we can manually delete the data associated with any specific part, or any specific group of parts. Just provide us with the part ID and we'll perform a hard delete for you. If you want the data removed, we'll remove it. Period. 
• If a Team administrator requests a Team to be deleted, we can manually delete all team data - including stored part data, tool libraries, and cut configs.

And additional details on data encryption are included below:

• Toolpath currently stores part data in S3 buckets which have built in data-at-rest encryption.
• Toolpath does not currently apply any additional encryption beyond those S3 buckets.
• None of the tool data stored in Toolpath, or general account information like email addresses, is encrypted.
• Toolpaths development and production environments are separate, so that our developers do not have casual access to user data. Our policy is that customer parts can only be accessed with customer permission. 
• We are currently starting a project to upgrade that whole system to be FedRamp certifiable.