Ownership, Licensing and Accountability

The human who merges is accountable

An agent is a tool; the person who accepts and merges the change owns it, including bugs, security flaws and outages. Practically: review as if a colleague wrote it; do not merge what you do not understand (ask the agent to explain, simplify, or add tests until you do); keep authorship and responsibility visible in commits and PRs. On licensing and provenance, models can reproduce patterns from training data, so follow your organisation's policy: some restrict which tools are allowed, require recording AI assistance, or scan for large copied blocks, and license terms for AI-generated code vary by jurisdiction and are still evolving, so check with your legal team for commercial work. On confidentiality, know what leaves your machine (prompts and files may go to a provider), choose tools and settings consistent with your data policy, and keep secrets out of reach.

Understand before you merge

If you cannot explain the change to a colleague, ask for simplification or more tests first.

Quick check: Who is accountable for a bug in code an agent wrote and you merged?

• You and your team, as the people who accepted the change
• Only the AI tool
• Nobody
• The compiler