Software developed in regulated industries — defence, medical devices, critical infrastructure — must satisfy requirements that change, sometimes abruptly, as standards are updated or new legislation takes effect. A long-running development project can find itself out of compliance mid-build, with the cost of correction scaling with how late the divergence is caught.

The project asks: how can agentic AI be designed to hold regulatory and standards context as a live constraint throughout a development project, and how should it respond when that context changes? The challenge has two parts — encoding the relevant regulatory requirements in a form the agent can reason over, and designing the agent's behaviour when a regulatory change conflicts with work already done or in progress. The team would select a specific regulatory domain (for example, IEC 62443 for industrial cybersecurity or IEC 62304 for medical device software) and prototype an agentic system that monitors development artefacts against the selected standard, flags deviations, and generates concrete remediation suggestions.

The prototype would also demonstrate a simulated regulatory update scenario and show how the agent propagates the change through its compliance model. The deliverable is a working prototype of a regulation-aware development agent, with a documented evaluation on a representative set of compliance scenarios and a design pattern for integrating live regulatory context into long-running software projects.