Code of Conduct clientApprovedApprovedApprovedApprovedConditionalConditionalConditionalConditionalConditionalNot permitted
AstraZeneca
Open — With Guardrails
AstraZeneca has published its own AI ethics framework rather than relying only on contract clauses — they expect partners to work within five principles: being explainable and transparent, fair, accountable, human-centric, and private and secure. It's a genuinely permissive account as long as you can show your working.
Key requirement — AstraZeneca's AI Code of Conduct is built around five principles — explainability, fairness, accountability, human-centricity, and privacy/security. This is a relatively permissive client; the ask is to document your AI use well enough to demonstrate you're meeting those principles, not to avoid AI.
Permitted
- AI tools for internal efficiency with HITL
- Ideation, research, drafting — with human review
- LynxAI (follows AZ's published AI Code of Conduct)
Not permitted
- AI use that skips AstraZeneca's published AI Code of Conduct
- Any output used without human-in-the-loop review
- Medical content bypassing MLR review
Tool-by-tool approval
LynxAI
HITL required
AVA
Enterprise; HITL required
Microsoft Copilot
Enterprise; HITL required
NotebookLM
Enterprise Google; HITL required
Vermeer
HITL; licensing compliance
Vurvey
HITL; verify with client
ElevenLabs
HITL; verify with client
HeyGen
HITL; verify with client
Runway
HITL; verify with client
Consumer AI
Not approved for client use