Client Guidelines Guide
Havas Baseline
For clients who do not yet have specific AI clauses in their contract (including Novo Nordisk, Teva, UCB, and any client not listed in this guide).
Approved tools
LynxAI (and its underlying models) only. No personal or consumer-tier AI tools for client work.
Permitted uses
- Internal efficiency tasks (drafting, research, summarising)
- Ideation and concept generation (with human editing)
- Translation and localisation support (with native speaker review)
- Data analysis of non-confidential datasets
Requires human review (HITL mandatory)
- All content that goes to the client or into production
- Any factual claims, statistics, or citations
- Medical, clinical, or regulatory content — medical review required regardless of AI involvement
Never permitted
- Processing personal data (PII) through AI without explicit consent and legal basis
- Using AI to fabricate quotes, data, or references
- Submitting AI-only outputs to MLR without disclosure and human review
When to escalate
If you're unsure whether a particular use case is covered by these guidelines, ask the AI team before proceeding. Escalate to the account lead if there's any question about client sensitivities.