Trust / Responsible AI
AI governance is part of the rebuild.
We're adopting AI in the products we build and in the way our teams work. Moving faster matters. So does being deliberate about the risks, the data, and the people affected by our decisions.
Why I'm writing about this
I'm leading Living Security through a transformation toward becoming AI-native. On this site, I'm sharing the decisions, experiments, and lessons along the way. Governance belongs in that story, not in a footnote.
AI introduces real questions. What information is appropriate to share with a tool? Where do we need human review? How do we evaluate an output before we trust it? Who owns the outcome when an automated process gets something wrong?
Those questions deserve clear answers as the technology and our use of it evolve.
The principles guiding my thinking
- Accountability. AI can assist with work and decisions, but people remain responsible for outcomes.
- Security and privacy. Data sensitivity, access, confidentiality, and vendor practices should shape how AI is used.
- Purpose. We should use AI to solve meaningful problems, not just because the technology is available.
- Testing and oversight. The higher the potential impact, the more important evaluation, monitoring, and a way to intervene become.
- Transparency. People should understand when AI materially shapes the experiences and decisions that affect them.
- Learning. Practices need to improve as tools, risks, and evidence change.
AI in our products
As we explore and build AI-enabled capabilities, customer trust has to be part of the product conversation. That means asking how features are evaluated, what data they use, where people retain control, and how limitations are communicated. Specific product capabilities and commitments should be understood through Living Security's official product and security materials.
AI across our teams
We're also changing how work gets done. I'm interested in what happens when people can build, experiment, and work with agents across functions. Speed is useful only if teams still understand the work, exercise judgment, and take responsibility for what ships.
What I'm still working through
How do we set appropriate review levels for different kinds of AI-assisted work? How do we assess third-party tools? What should we measure to know an AI-enabled workflow is working safely and well? How do we make it easy to raise concerns?
I won't pretend every answer is settled. I'll share what I can as the work develops, without disclosing confidential customer or company information.