Back to Handbook

Governance · Advanced

Security, Risk & Legal

Safety First

Understanding the unique risks of AI-generated code, including Hallucinations and Copyright.

01

Package Hallucinations

Sometimes AI invents software libraries that don't exist. Hackers can create malicious versions of these 'fake' libraries to steal your data.

Security checkpoint

Verify generated packages, permissions, and sensitive data paths before shipping.

02

Shadow AI

Be careful pasting private company code into public AI models. It might be used to train future models, leaking your secrets.

03

Transparency & Compliance

In the age of AI, trust is your most valuable asset. Users want to know how the 'Magic' works.

  • System Cards: Publish a document explaining which models you use and how you handle data.
  • GDPR/EU AI Act: Ensure your AI features allow users to 'Opt Out' and delete their data.
  • Human in the Loop: For high-stakes actions (payments, emails), always require user confirmation before the Agent acts.

04

Copyright & IP Strategy

The legal landscape of AI is shifting. Understand ownership.

  • Output Ownership: In many jurisdictions, AI-generated code cannot be copyrighted. Human modification is key to claiming ownership.
  • Trademarking Vibe: You can trademark the 'Brand' (Name, Logo) even if the code is AI-generated.
  • License Laundering: Be careful not to let the AI output GPL code into a proprietary project. Use .cursorrules to enforce 'MIT Only'.