Overview
ChatLemur is committed to developing and deploying artificial intelligence responsibly. We align our AI assistant, Lenny, with Australia's voluntary AI Ethics Principles developed by the Department of Industry, Science and Resources, as well as CSIRO's Responsible AI guidelines.
This framework outlines how we implement each principle in practice and our ongoing commitment to ethical AI development as an Australian company.
Australia's 8 AI Ethics Principles
The following principles guide all AI development at ChatLemur. Each principle includes our specific implementation approach for Lenny AI.
1 Human, Societal and Environmental Wellbeing
Principle: AI systems should benefit individuals, society and the environment.
Our Implementation:
- Lenny is designed to augment human capabilities, not replace human connection
- Built with accessibility at the core (VoiceOver, keyboard navigation, screen reader support)
- Powered by efficient cloud infrastructure with carbon-neutral data centres
- Encourages human-to-human chat alongside AI assistance
- Never promotes harmful content, violence, or illegal activities
2 Human-Centred Values
Principle: AI systems should respect human rights, diversity, and individual autonomy.
Our Implementation:
- Users maintain full control over AI interactions (can disable AI, use @human mode)
- Respects diverse communication styles and cultural backgrounds
- Never makes decisions on behalf of users without explicit consent
- Created by a blind developer with accessibility as a fundamental right, not an afterthought
- Supports voice input for users with different abilities
3 Fairness
Principle: AI systems should be inclusive and accessible, and should not discriminate unfairly.
Our Implementation:
- Lenny does not discriminate based on gender, race, age, disability, or background
- Equal service quality for all pricing tiers (AI quality is consistent)
- Regular bias audits of AI responses and model outputs
- WCAG 2.1 AA accessibility compliance for all users
- Inclusive language guidelines built into Lenny's training
4 Privacy Protection and Security
Principle: AI systems should respect and uphold privacy rights and data protection.
Our Implementation:
- All data encrypted in transit (TLS 1.3) and at rest (AES-256)
- No conversation data sold to third parties
- Conversations are not used to train external AI models
- Compliant with Australian Privacy Principles (APPs)
- Users can request data deletion at any time
- Minimal data collection - only what's necessary for service
5 Reliability and Safety
Principle: AI systems should reliably operate in accordance with their intended purpose.
Our Implementation:
- Robust fallback systems if AI services are unavailable
- Clear boundaries on what Lenny can and cannot do
- Refuses to provide medical, legal, or financial advice
- 99.9% uptime SLA with graceful degradation
- Continuous monitoring for hallucinations and factual errors
- Emergency stop capabilities for administrators
6 Transparency and Explainability
Principle: There should be transparency about when AI is being used and how it makes decisions.
Our Implementation:
- Lenny always identifies as an AI assistant (never pretends to be human)
- Clear visual indicators when AI is responding vs humans
- Lenny explains reasoning when asked "why did you say that?"
- Open documentation about AI capabilities and limitations
- Discloses when information might be uncertain or speculative
7 Contestability
Principle: When AI significantly impacts a person, they should be able to challenge it.
Our Implementation:
- Users can report incorrect or inappropriate AI responses
- Feedback mechanism: "Was this response helpful?" on every AI message
- Human review process for escalated complaints
- Ability to request human-only interaction mode
- Appeals process for account-related AI decisions
- Contact: ethics@chatlemur.com
8 Accountability
Principle: Those responsible for AI should be identifiable and accountable.
Our Implementation:
- ChatLemur Pty Ltd (ABN: 26 994 122 501) is accountable for Lenny's behaviour
- Founder: Sarah Mitchell (AI Ethics Officer)
- Regular internal ethics reviews and audits
- Published incident response procedures
- Liability insurance covering AI-related issues
- Annual transparency report on AI operations
How Lenny AI is Designed Ethically
Bias Mitigation Measures
- Base model (Llama 3.3 via Groq) selected for balanced training data
- System prompts explicitly instruct against discriminatory responses
- Regular testing with diverse demographic scenarios
- Community feedback incorporated into bias detection
- No reinforcement of stereotypes in responses
Human Oversight Mechanisms
- Administrators can pause AI responses globally
- Content moderation layer filters harmful outputs
- Rate limiting prevents AI abuse
- Audit logs of all AI interactions (anonymised)
- Human review queue for flagged responses
AI Decision Transparency
- Lenny explains its reasoning when asked
- Confidence levels indicated for uncertain information
- Sources cited when providing factual claims
- Clear distinction between facts and opinions
How to Contest AI Outputs
If you believe Lenny has provided incorrect, harmful, or biased information, you have multiple avenues to contest and seek resolution:
- In-App Feedback: Click "Report" on any AI message to flag it for review
- Email: Contact ethics@chatlemur.com with details
- Human Mode: Use @human to bypass AI entirely
- Escalation: Request human review of your case within 48 hours
All contestation requests are logged, reviewed by a human team member, and responded to within 5 business days.
Accountability for AI Behaviour
| Role | Responsibility | Contact |
|---|---|---|
| AI Ethics Officer | Overall AI ethics governance and policy | Sarah Mitchell |
| Technical Lead | AI model selection, safety controls, monitoring | Engineering Team |
| Support Team | User complaints, feedback processing | support@chatlemur.com |
| Legal Entity | Corporate accountability, compliance | ChatLemur Pty Ltd |
AI Training Data Disclosure
Transparency about what data trains our AI is essential. Here's what you should know:
| Data Type | Used for Training? | Details |
|---|---|---|
| User conversations | No | Your chats are NEVER used to train AI models |
| Base model training | N/A (third-party) | Llama 3.3 trained by Meta on public data |
| System prompts | Yes (by us) | ChatLemur-specific personality and guidelines |
| Knowledge base | Yes (by us) | Product documentation, help content |
| Feedback data | Aggregated only | Anonymous patterns to improve responses |
Environmental Impact Statement
We recognise that AI systems have environmental costs. ChatLemur is committed to minimising our carbon footprint:
- Efficient Models: We use Groq's LPU infrastructure, which is significantly more energy-efficient than traditional GPU clusters
- Cloud Provider: Hosted on Fly.io with renewable energy commitments
- Model Selection: Chose Llama 3.3 for its efficiency-to-quality ratio
- Caching: Intelligent caching reduces redundant AI calls
- Local Options: agentic-brain supports local LLMs for reduced cloud dependency
We commit to publishing annual environmental impact assessments as our usage scales.
CSIRO Responsible AI Alignment
ChatLemur aligns with CSIRO's Responsible AI Network guidelines, which complement Australia's national AI Ethics Principles:
- Risk Assessment: Regular AI risk assessments conducted
- Governance: Clear AI governance structure in place
- Monitoring: Continuous monitoring of AI behaviour and outputs
- Documentation: Comprehensive documentation of AI systems
- Stakeholder Engagement: User feedback integrated into development
⚠️ Future-Proofing for Mandatory Regulations
While Australia's AI Ethics Principles are currently voluntary, mandatory AI regulations are being considered. ChatLemur is prepared for potential requirements including:
- Mandatory AI impact assessments
- AI system registration requirements
- Algorithmic transparency obligations
- High-risk AI classification compliance
- Cross-border data flow restrictions
We actively monitor regulatory developments through the Department of Industry, Science and Resources and will update our practices as requirements evolve.
Our Commitment
ChatLemur believes that ethical AI is not just a compliance checkbox—it's fundamental to building technology that genuinely helps people. As an Australian company founded by a developer who relies on accessible technology daily, we understand that AI must serve all users equitably.
We commit to continuous improvement of our AI ethics practices and welcome community feedback on how we can do better.
Questions or Concerns?
We welcome dialogue about our AI ethics practices. Contact us at:
Email: ethics@chatlemur.com
General: joseph.webber@gmail.com