AI Model Transparency Statement
How our models work, what they use, and their known limitations.
Last updated July 12, 2026
1. Purpose
Haruspex (“Haruspex”, “we”, “our”, or “us”) is committed to providing transparent information about how artificial intelligence is used throughout our platform.
This AI Model Transparency Statement explains:
- how our AI systems operate,
- what models may be used,
- where AI-generated content comes from,
- the limitations of AI outputs,
- how human judgment should be applied, and
- our commitment to responsible AI development.
This document should be read together with our:
- Terms of Service
- Privacy Policy
- AI & Investment Disclosure
- Risk Disclosure Statement
- Acceptable Use Policy
2. AI-First Platform
Haruspex is an AI-powered investment intelligence platform.
Most platform functionality is generated, assisted, or enhanced by artificial intelligence, including but not limited to:
- investment research
- company analysis
- macroeconomic analysis
- sector analysis
- risk evaluation
- portfolio insights
- market summaries
- earnings interpretation
- document summarization
- news synthesis
- natural language conversations
- explanation generation
- educational content
Users should assume that any analytical output displayed by Haruspex has been generated, at least in part, by artificial intelligence.
3. Multiple AI Models
Haruspex is model-agnostic.
We may use one or more AI providers depending on the task being performed.
These may include:
- proprietary Haruspex models
- commercial foundation models
- open-source models
- specialized reasoning models
- embedding models
- ranking models
- summarization models
- retrieval models
The specific model used for any request may change without prior notice.
4. Dynamic Model Selection
Different requests may be processed by different AI systems.
Model selection may depend on factors including:
- reasoning complexity
- latency requirements
- context length
- language
- cost optimization
- safety requirements
- reliability
- availability
- infrastructure health
Users should not assume that every response is produced by the same model.
5. Retrieval-Augmented Generation (RAG)
Some Haruspex responses are generated using Retrieval-Augmented Generation (RAG).
This means AI models may combine:
- user prompts
- internal knowledge
- structured financial data
- market data
- public information
- company disclosures
- news
- historical information
- cached analysis
The underlying model does not necessarily “know” all information independently.
Instead, information may be retrieved dynamically before a response is generated.
6. External Data Sources
Haruspex may incorporate information from third-party providers, including:
- financial market data
- public company filings
- news providers
- macroeconomic datasets
- economic indicators
- publicly available information
- licensed datasets
- user-provided information
Availability of data may vary by jurisdiction, provider availability, licensing restrictions, or technical limitations.
7. AI Generated Content
Unless explicitly stated otherwise, content produced through Haruspex should be considered AI-generated.
Examples include:
- reports
- summaries
- forecasts
- explanations
- scoring
- rankings
- investment theses
- opportunity analysis
- risk analysis
- portfolio observations
- chat responses
8. AI Does Not Possess Human Judgment
AI models:
- do not think like humans,
- do not possess consciousness,
- do not understand context perfectly,
- do not independently verify every statement,
- may generate incorrect conclusions,
- may overlook important facts,
- may misunderstand prompts.
Outputs represent probabilistic predictions rather than verified facts.
9. Hallucinations
Artificial intelligence systems can generate inaccurate information (“hallucinations”).
These may include:
- fabricated citations
- incorrect financial figures
- outdated information
- nonexistent events
- inaccurate reasoning
- incorrect calculations
- false relationships
- mistaken assumptions
Users remain responsible for independently verifying important information before making decisions.
10. Investment Analysis Limitations
Although Haruspex specializes in investment intelligence, AI analysis should never be interpreted as:
- financial advice,
- investment advice,
- legal advice,
- accounting advice,
- tax advice,
- fiduciary guidance,
- personalized recommendations.
Investment decisions should always involve independent research and professional judgment where appropriate.
11. Model Updates
Haruspex continuously improves its AI systems.
Improvements may include:
- newer models
- different providers
- updated prompts
- better retrieval systems
- improved reasoning
- improved safety systems
- expanded datasets
- infrastructure changes
As a result, identical prompts submitted at different times may produce different responses.
12. No Guaranteed Consistency
AI systems are probabilistic.
Responses may vary because of:
- model updates
- retrieval updates
- market changes
- prompt wording
- context differences
- system improvements
Users should not rely on identical outputs across different sessions.
13. Bias
AI systems may inherit biases present in:
- training data
- public information
- financial reporting
- historical datasets
- internet content
- market narratives
Haruspex actively works to reduce harmful bias but cannot guarantee its complete elimination.
14. Safety Controls
Haruspex implements safeguards intended to improve the safety and reliability of AI outputs.
These may include:
- prompt filtering
- safety classifiers
- moderation systems
- validation rules
- retrieval constraints
- output filtering
- abuse detection
- rate limiting
These controls reduce risk but cannot eliminate all incorrect or harmful outputs.
15. User Responsibility
Users are responsible for:
- reviewing AI outputs critically,
- verifying material facts,
- checking calculations,
- validating market information,
- consulting primary sources,
- making independent investment decisions.
AI should support decision-making, not replace it.
16. Transparency About Confidence
Where technically feasible, Haruspex may communicate uncertainty through:
- confidence indicators
- probability estimates
- supporting evidence
- cited sources
- alternative interpretations
- identified assumptions
Not all responses will include confidence metrics.
17. Human Oversight
Haruspex may review certain AI outputs for purposes including:
- quality improvement
- abuse prevention
- system debugging
- safety evaluation
- legal compliance
- service reliability
Such reviews are conducted in accordance with our Privacy Policy and applicable internal access controls.
18. User Feedback
Users are encouraged to report outputs that appear:
- inaccurate
- misleading
- harmful
- biased
- incomplete
- unsafe
- offensive
- technically incorrect
Feedback helps improve future versions of the platform.
19. Availability of Models
Haruspex does not guarantee:
- availability of any specific AI model,
- continued support for any provider,
- permanent access to particular features,
- identical model behavior over time.
Models may be replaced, upgraded, retired, or temporarily unavailable without notice.
20. Intellectual Property
AI-generated outputs may incorporate statistical patterns learned during model training but are not intended to reproduce copyrighted works.
Users remain responsible for ensuring that their use of generated content complies with applicable laws and the rights of third parties.
21. Changes to This Statement
We may update this AI Model Transparency Statement from time to time to reflect:
- new AI capabilities,
- changes to model providers,
- regulatory developments,
- industry standards,
- platform improvements,
- safety enhancements.
The latest version will be published on the Haruspex website, and the Effective Date will be updated accordingly.
22. Contact
Questions regarding this AI Model Transparency Statement may be directed to:
Haruspex
Website: https://haruspex.guru
Support: support@haruspex.guru