[Phase 0.3.1] Create AI_PRINCIPLES.md — Core AI safety principles

[richard-devbot]

Summary

Closes #11

Creates AI_PRINCIPLES.md at the project root defining the 6 core AI safety principles governing all Operator-Use development:

1. Least Privilege — minimum permissions for the current task
2. Human Oversight — no irreversible action without confirmation
3. Transparency — explain before acting
4. Containment — bounded, sandboxed, reversible actions
5. Privacy by Default — minimum data access/storage
6. Fail Safe — stop and ask when uncertain

Each principle includes definition, Operator-Use-specific implementation guidance, compliant/non-compliant examples, and testing criteria.

Also adds a reference link in README.md under the Security section.

Test plan

• All 6 principles present with all required subsections (definition, implementation guidance, compliant example, non-compliant example, testing criteria)
• References section present with all 5 external standards
• README.md Security section links to AI_PRINCIPLES.md
• Committed on richardson/phase0-ai-principles, not main
• PR targets CursorTouch/Operator-Use main and references issue [Phase 0.3.1] Create AI_PRINCIPLES.md — Core AI safety principles #11

References

• NIST AI RMF (AI 100-1)
• EU AI Act
• Anthropic Core Views on AI Safety
• IEEE 7000
• Google AI Principles

🤖 Generated with Claude Code