Best Practices for AI-Assisted Code Generation Tools (Like CoPilot)
Talkat Open Source Summit Europe 2023 · on Wednesday, 2023-09-20· at 12:20· Room 3H (3JAUREGIA - Floor 3), Abandoibarra Etorb., 4, 48011 Bilbao, Bizkaia, Spain·
summary
AI-assisted code generation causes five concrete risks: copyrighted memorized outputs, trade secret leakage if models are fine-tuned on private repos, security regressions from learned vulnerable patterns, unclear ownership under current US copyright guidance, and skill atrophy if humans stop editing. Concrete mitigations: mark AI-generated files so scans focus on diffs, adopt test-driven development to constrain and detect bad or copied code, and run weekly snippet scanning with human annotations; GitHub/Copilot and CodeWhisperer show attribution/indemnity limits.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 4:06Copyright Infringement
- 8:40Proprietary Code
- 10:31Security
- 11:34Copyright
- 16:35Too Far Away
- 18:13Best Practices
- 18:43Best Practice 1
- 19:42Best Practice 2
- 23:30Best Practice 3
- 34:52Summary
related talks
43mBest Practices for AI-Assisted Code Generation Tools (Like CoPilot)
44mAI Code Generation - Benefits, Risks and Mitigation Controls
33mAI assisted software delivery
30mWhat Are the Security Risks Associated with AI-Generated Code? 🤔
19mWith great power comes great responsibility in code quality
51mAI-Assisted Coding: The Future; Between Challenges And Benefits
38mThe Cognitive Cost of AI-Assisted Open Source
23mAI-Generated Code: Who Is Responsible for the Bugs?