What happens when AI-generated code ships faster than humans can properly review it, and who takes the blame when something breaks?

In this episode of AI at Work, I sit down with Cameron Etezadi, Chief Technology Officer at LaunchDarkly, to tackle one of the most uncomfortable questions facing modern software teams. As developers increasingly rely on AI coding assistants, copilots, and public LLMs to accelerate delivery, organizations are finding themselves caught between productivity gains and growing governance risks.

Cameron explains why “Shadow AI” has become the modern evolution of Shadow IT, and why the stakes are far higher when AI-generated code is moving directly into production systems. We explore how engineering teams are balancing innovation with accountability, why runtime controls and kill switches are becoming essential in AI-native software development, and how organizations are struggling to maintain visibility into code generated by autonomous systems. Cameron also explains why he believes many companies are unknowingly exposing intellectual property, customer trust, and compliance obligations through careless AI use.

The conversation also examines how the EU AI Act and Product Liability Directive could reshape software development globally. Cameron argues that organizations deploying AI-generated code are now effectively treated as manufacturers under emerging regulations, with accountability resting firmly on businesses shipping software, not the AI vendors creating the tools. From governance gaps and auditability concerns to token economics and developer productivity metrics, this discussion explores the operational realities behind the AI hype cycle.

We also discuss why faster code does not automatically mean safer software, the hidden costs of AI-generated rework, and how some organizations are already spending more time fixing AI-assisted production issues than they expected. Cameron shares practical advice for boards, CISOs, and DevOps leaders on what questions they should be asking today before AI governance problems become tomorrow’s security incidents.

If your organization is experimenting with AI-assisted development, this conversation offers a valuable reality check on where the risks are emerging, how the rules are changing, and why accountability still matters in an increasingly automated world.