Question: Who owns AI-generated code?

Presentation from TECH LEADERS FORUM 2026, Riga (https://ninebrains.eu/tech-leaders-forum/)

Question: Who owns AI-generated code? Where should responsibility for AI-generated code remain with humans?

Presentation is deliberately short — it's a discussion starter on the topic. The real value comes from peers exchanging experience on similar challenges.

Transcript

1. Question Who owns AI-generated code? Alexey Buzdin Director of Engineering @Wandoo Finance · Sep 2026
2. My conversation with an Engineer 02
3. My conversation with an Engineer Me “I see you have submitted a PR. It was approved a week ago, but it's not merged. Can you merge it please, or what's the status?” 03
4. My conversation with an Engineer “I see you have submitted a PR. It was approved a week ago, but it's not merged. Can you merge it please, or what's the status?” Engineer “I can merge it, no problem, but I would like a QA specialist to test the code before I merge, since I don't know how the code change works. AI wrote it for me.” 04
6. Don't be a MeatProxy meatproxy.me 06
7. The reframe Treat AI-generated code as one coming from a fellow Junior engineer 07
8. Who owns Junior's code? 08
9. How to teach a Junior engineer to grow? 1 Explain Definition of Done / Success Metrics 4 Provide ways to isolate WIP (e.g. Feature Flags) 7 Teach how to test changes locally before a PR 2 Break down the Epic into small, manageable chunks together 5 Teach how to deliver atomic changes 8 Review Pull Requests and provide feedback 3 Explain team coding conventions 6 Have an automated regression test suite 9 Monitor changes post-launch 09
10. How to trust your generated code? 1 of 4 Move focus from reviewing code to reviewing specifications to the AI (SDD / PRD) But… “Why would I wait for a review of the spec if I can just generate the code?” 10
11. How to trust your generated code? 2 of 4 Feed team coding conventions and ADRs to the AI agent (AGENTS.md or a custom Skill) and make sure Pull Request comments improve team's AI memory bank. But… “I used the Skill, sometimes it didn't work and used too many tokens so I disabled it.” 11
12. How to trust your generated code? 3 of 4 Make sure your AI agent can run the product locally and use an automated regression test suite to loop back until the change is done and tests are passing. But… “If it takes hours or days to merge code changes and deploy them, the momentum is lost.” 12
13. How to trust your generated code? 4 of 4 Use a different AI agent to review the Pull Request for correctness and backward compatibility (Claude vs Codex, Sonnet vs Opus). But… “How much would this cost?” 13
14. Thank you! Questions for debate 1 Where does responsibility for AI-generated code sit at your company today, and who actually enforces it? 2 What's your review process for AI-generated code, and is it different from reviewing humanwritten code? 3 Have you had an incident or near-miss caused by AI-generated code? What changed afterward? 4 What would you never let AI touch without a human sign-off, and why that line specifically? 5 If you had to defend your current code-ownership policy to a regulator tomorrow, where would it fall apart? 14