Full disclosure of AI providers, models, and data processing practices.
Trust Center
AI you can trust
Trust shapes our AI practices. Every solution we build is grounded in transparency and clear governance.
AI Principles
Transparency
Data Protection
Zero data retention with AI providers. Customer data is never used for public model training.
Quality Control
Human oversight, validation processes, and continuous accuracy monitoring.
Bias Mitigation
Regular testing and auditing to identify and eliminate algorithmic bias.
User Control
Granular settings to control AI features and data processing preferences.
Product-specific AI usage
We utilize a sophisticated stack of large language models (LLMs), small language models (SLMs), and proprietary machine learning models to power the Lusha platform.
Search & information extraction
- Models: Flex NER, Social Posts NER, Search NG Extractor, Search NG Normalizers
- Type: LLM, SLM, and proprietary Models
- Customer data used: No
- Purpose: Powers high-precision natural language search, intent understanding, and the extraction of entities from professional social content
Data normalization & categorization
- Models: Job Title Normalization, Industries (including SIC/NAICS), Specialities (Grouping), and Name Normalization
- Type: LLM, SLM, and Model + LLM Fallback
- Customer data used: No
- Purpose: Standardizes fragmented data into clean, searchable categories to ensure database consistency and accuracy
Business intelligence & technographics
- Models: Technographics, Company Keywords and Specialities, Revenue, and Competitors (WIP)
- Type: SLM and proprietary Models
- Customer data used: No
- Purpose: Identifies company tech stacks, keywords, financial brackets, and competitive landscapes to provide deep firmographic insights
Contact enrichment & validation
- Models: One Time Email, Reverse Lookup, and Seniority
- Type: LLM and proprietary Models
- Customer data used: No
- Purpose: Enhances contact profiles, validates email deliverability in real-time, and determines professional hierarchy
Conversational AI & analysis
- Models: Chat Coordinator, Chat Web Search, CI (Meeting Analysis), and Guardrails
- Type: LLM
- Customer data used: No
- Purpose: Facilitates intelligent chat interactions and meeting summaries while maintaining strict safety guardrails to prevent hallucinations or data leakage
Personalized recommendations
- Models: Recommended Searches, Two Tower Recommendations (WIP), and CRM predictions
- Type: Proprietary models
- Customer data used: Yes, customers can opt-out by reaching out to Lusha’s Support team
- Purpose: Analyzes usage patterns to suggest relevant contacts and companies tailored to customer needs
Account health & operations
- Models: Churn (Enterprise Accounts)
- Type: Proprietary model
- Customer data used: Lusha internal data only
- Purpose: Internal monitoring of enterprise account health to improve service delivery and customer success
Data security in AI processing
- All data encrypted in transit (TLS 1.3) to AI providers
- Zero customer data retention with AI providers
- No use of customer data for AI model training
- Contractual data protection agreements with all AI providers
- Secure API authentication and access controls
- Regular security audits of AI integrations
- Data minimization – only necessary data sent to AI systems
Quality controls & bias mitigation
Accuracy & quality
- Human review of AI-generated outputs
- Accuracy benchmarking and continuous monitoring
- A/B testing of AI improvements
- Feedback mechanisms for reporting issues
- Regular model performance evaluation
Bias testing & fairness
- Regular bias testing across demographic dimensions
- Fairness audits for recommendation algorithms
- Diverse training data to reduce bias
- Transparent documentation of limitations
- Continuous improvement based on feedback
User controls
- Account-level opt-out of all AI-powered features
- Granular controls for individual AI capabilities
- Clear labeling of which features use AI
- Data processing preferences management
- Transparency about AI vs. human-generated content
Governance & accountability
AI ethics & governance framework
- Cross-functional AI ethics committee
- Regular risk assessments for AI systems
- Clear accountability and decision-making processes
- Compliance with emerging AI regulations
TRUSTe Responsible AI Certification
Independent validation of our AI practices, data handling, transparency standards, and ethical governance. Lusha is one of the first companies in sales intelligence to achieve this certification.
ISO 42001
ISO 42001 AI Management System certification demonstrates comprehensive AI governance aligned with international standards.
Regular audits
- Quarterly internal reviews of AI systems and practices
- Annual third-party AI ethics audits
