WorkSeptember 14, 2026via TechCrunch AI
Microsoft’s new AI ‘code of conduct’ tells models not to hack systems or trick humans
Why it matters
Microsoft's published AI code of conduct signals a company-wide governance stance on model behavior and safety — a policy/cultural move relevant to how enterprises will govern AI deployment and risk, but lacks technical depth or industry-wide adoption data to drive immediate practitioner decisions.
Key signals
- Microsoft issues formal AI code of conduct for internal models
- Principles include: supporting humans (not replacing), accelerating human flourishing
- Specific safety constraints: prohibits hacking systems, deception
- Framed as governance/policy layer, not a technical capability or model release
- No third-party validation or adoption metrics provided
- Published Sept 2026
- Microsoft issued formal 'code of conduct' for AI models
- Constraints include: no hacking systems, no deceiving humans
- Principles include: supporting rather than replacing humans, accelerating human flourishing
- Published September 14, 2026
- Framed as safety implementation, not binding policy
The hook
Microsoft codifies AI safety rules: no hacking, no deception. But can principles survive at scale?
The code of conduct lays out general principles that Microsoft AI models should uphold — supporting humans rather than replacing them, for instance, and accelerating human flourishing — as well as specific safety constraints meant to implement those principles.