Artificial intelligence (AI) is transforming the way infrastructure is designed, built, and operated—and software from Bentley Systems is at the forefront of this change. Our applications help engineers and organizations unlock new levels of productivity, optimize decisions, and tap into the full value of their data. With these opportunities, however, comes a critical responsibility: ensuring that our AI is built on trust, transparency, and respect for our users’ data. For engineers, data is not just information; it’s intellectual property and a strategic asset.
At the 2023 Year in Infrastructure (YII) Conference, Bentley emphasized that users retain full access and control over their data stored and managed in Bentley products, including whether and how it’s used for AI training. Our users’ data is their data, always. This principle has guided our approach to developing infrastructure AI capabilities, empowering our users to confidently innovate, knowing their interests are safeguarded.
Bentley has adopted robust controls to support the training and development of AI models. Among those controls are two categories of AI models that we designed to respect user preferences and maintain our commitment to data stewardship:
- Bentley AI Models: We developed these models by using data that has been purchased and/or licensed for training purposes. Licensed data can include data that users and accounts have agreed to contribute for the purpose of developing AI models that benefit all Bentley users.
- Fit for Account Models: Bentley AI Models can be fine-tuned by users within their own instances. Data specific to each account is utilized solely for the benefit of that account and is not incorporated into broader model or product enhancements.
This approach ensures that users retain control over their data and can choose whether to contribute to broader AI model development or keep their data and model improvements specific to their own organization.
While these controls enable users to decide how their data is used, transparency remains essential. That’s why Bentley has developed the Data Agreement Registry (DAR), an auditing system to enhance visibility for users and accounts. DAR lets users monitor model deployment and performance, and retain control over how their data is being used. It offers detailed insights to support auditing of model performance, particularly for accounts tracking the benefits of using Bentley AI Models versus Fit for Account Models. With DAR, our users have clarity and control, reinforcing Bentley’s commitment to responsible, trustworthy AI.
Our approach to data stewardship is foundational to Bentley’s AI practices and is supported through our global risk management programs. Bentley’s privacy and security programs are benchmarked against global standards, including ISO 27001, ISO 27701, and SOC 2. Similarly, our AI practices are being aligned to international standards, including ISO 42001 and NIST’s AI Risk Management Framework. But our goal isn’t just alignment—it’s leadership. Bentley is setting a higher bar for ethical, transparent AI in infrastructure.
As AI regulations and practices evolve, Bentley will continue to enhance DAR and our governance framework to meet emerging standards. Our commitment to data stewardship in the age of AI is clear: Bentley protects intellectual property and gives users full access and control over their data, including how it’s used for AI training.
