How MacroFab uses artificial intelligence — and the rules we hold ourselves to in order to protect your intellectual property and data.
As the pace of technology rapidly increases, AI has enabled significant advancements in software development, time to deliver solutions, and the costs of service delivery. In this way, AI has become essential for most companies to meet the demands of customers and keep pace with the modern world. Unfortunately, the rush to implement AI-driven solutions has in many cases not been met with a clear strategy for protecting customer information.
At MacroFab, we understand both the importance of AI and the risk it imposes on the protection of intellectual property and quality outcomes — especially for the most important products being manufactured in the United States. For this reason, we have implemented a focused policy for the use of AI technology both within the MacroFab Platform and within the back-office tools used by our employees. Here we outline the core elements of that policy.
FabIQ is a powerful tool designed to save engineers time and money and to reduce risk in electronics manufacturing. It underpins many of the features in the MacroFab Platform. FabIQ is not a single AI system, but a collection of distinct intelligence tools, each operating under the following requirements:
Every AI-driven feature trained with proprietary data held by MacroFab must be scoped to the design it is currently operating against in the most restrictive context, and may not make recommendations that would create the potential for leakage of information across customer boundaries. Customer isolation across data boundaries must be maintained and must be verifiable.
MacroFab manages data about in-process and post-process defects across tens of thousands of orders. This capability is substantively powerful when used correctly, but implementation details can create significant risk of cross-customer leakage — and our rules are designed to prevent this.
Predictive defect analysis can identify components or design attributes with a higher risk of manufacturing defects, but must limit its guidance to high-level suggestions for improvements in the context of the current design and/or the component datasheet. For example, FabIQ can flag incorrect geometry in a component footprint and suggest "Reduce pads by 33% or reduce paste deposition by 50%" or "The datasheet recommends this specific footprint for this component." FabIQ cannot recommend a landing pattern that does not exist in the datasheet, as doing so would create the opportunity to leak data from another customer's design.
AI has become essential in all back-office processes — from context generation mechanisms that summarize key points of long-running email threads, to financial analysis and document generation. To assist in protecting customer information and reducing risk, MacroFab has established the following policies for AI within back-office tools:
When dealing with export-controlled information (ITAR or EAR) or Controlled Unclassified Information (CUI), the utilization of any AI feature or tool shall not result in the crossing of ITAR or CUI boundaries. Any such feature shall be automatically disabled.
Compliance with all applicable law and regulations with respect to such information shall be enforced at all times.