Legal and disclosure information
Methodology and AI Limitations
Kerdos AI is an AI-assisted financial research and analytics tool. It uses quantitative signals, market data, public data, company data, news and sentiment data where available, statistical models, machine-learning models, and AI-generated explanations to organise market research.
Kerdos AI outputs are provided for informational and educational research purposes only. They are not personalised investment advice, portfolio management, brokerage, order execution, custody, tax advice, or legal advice. Nothing on Kerdos AI should be understood as a recommendation, instruction, solicitation, or offer to buy, sell, hold, subscribe for, or trade any financial instrument.
Current Methodology Summary
- Data frequency
- End-of-day, with some delayed and mixed-frequency sources
- Main source categories
- Market price & volume data, Fundamentals & company data, News & sentiment data (where available), Macroeconomic & market-context data, Technical & quantitative indicators
- Ranking type
- Relative universe ranking
- Forecast type
- Scenario-based model output
- Regime type
- Market-context classification
- Last methodology update
- 7 June 2026
- Methodology version
- v1.0
How Outputs Are Produced
Kerdos AI may combine several types of information, depending on availability and product configuration, including:
- market price and volume data;
- technical and quantitative indicators;
- fundamental and company-related data;
- sector, industry, and exchange-level information;
- macroeconomic and market-context data;
- public news and sentiment data where available;
- historical patterns and statistical features;
- AI-generated summaries and explanations;
- internal model outputs such as rankings, research signals, confidence indicators, forecast scenarios, and regime classifications.
Not every output uses every data source. Data availability, quality, freshness, and coverage can vary by instrument, issuer, exchange, market, provider, and time period.
Outputs are probabilistic, uncertain, and model-dependent. They can change when new data becomes available, when assumptions change, when market conditions change, or when Kerdos AI updates its models or methodology.
Research Signals
Kerdos AI may classify instruments into research signal categories. These categories are model-generated research classifications only. They are not buy, sell, hold, or trade instructions.
Signal categories may include:
- Very Positive Signal: a model-generated research category indicating strongly positive evidence within the selected methodology.
- Positive Signal: a model-generated research category indicating positive evidence within the selected methodology.
- Neutral Signal: a model-generated research category indicating mixed, limited, unclear, or balanced evidence.
- Negative Signal: a model-generated research category indicating negative evidence within the selected methodology.
- Very Negative Signal: a model-generated research category indicating strongly negative evidence within the selected methodology.
A positive signal does not mean that an instrument is suitable for any user. A negative signal does not mean that an instrument must be sold or avoided. Signal labels are research classifications and should be reviewed together with the underlying assumptions, source context, risk factors, and the user’s own independent research.
Universe Ranking
Kerdos AI may rank instruments within a selected research universe. A research universe may be defined by factors such as exchange, region, sector, market capitalisation, liquidity, data availability, instrument type, or other filters.
Universe rankings are relative model outputs. A higher-ranked instrument is ranked higher only within the selected universe and selected methodology at a particular point in time. It does not mean that the instrument is objectively the best investment, suitable for any user, or expected to produce a profit.
Important limitations of universe ranking include:
- Rankings depend on the instruments included in the universe.
- Instruments excluded from the universe are not considered in the ranking.
- Rankings may change when the universe, filters, data, model version, or methodology changes.
- A rank can be affected by missing data, stale data, outliers, liquidity constraints, survivorship bias, or provider coverage.
- Closely ranked instruments may not be meaningfully different from each other.
- A high rank does not remove market, liquidity, valuation, company-specific, sector, currency, or macroeconomic risk.
Universe rankings should be treated as a research prioritisation tool, not as a recommendation to buy, sell, hold, or trade.
Regime Calculation
Kerdos AI may estimate or classify market regimes, sector regimes, instrument-level regimes, volatility regimes, trend regimes, liquidity regimes, sentiment regimes, or other market-context states.
Regime calculations are model-generated classifications based on selected data and assumptions. They are not certain descriptions of the market and are not reliable predictions of future conditions.
Important limitations of regime calculations include:
- Regime labels may lag fast-moving markets.
- Regime labels may change when new data becomes available.
- Different models can classify the same market differently.
- Market regimes can shift suddenly due to news, earnings, macroeconomic events, central-bank decisions, geopolitical events, liquidity shocks, or regulatory changes.
- A regime classification may be incorrect, incomplete, unstable, or too broad for a specific instrument.
- A regime label does not mean that any specific trade, portfolio allocation, or investment strategy is suitable or appropriate.
Regime calculations are intended to provide research context. They should not be used as standalone trading signals or as the sole basis for investment decisions.
Forecast Horizons and Scenario Price Targets
Forecast horizons describe the time period that a model scenario is intended to analyse. Kerdos AI may produce short-term, medium-term, or longer-term research scenarios depending on the available product features and data.
Scenario price targets are hypothetical model outputs based on available data and assumptions. They are not guarantees, not promises, and not reliable predictions of future prices.
Where Kerdos AI displays upside, downside, or base scenarios, these should be understood as model-generated scenarios, not as expected outcomes or investment recommendations.
Forecasts and scenario price targets may be unreliable during unusual market conditions, low-liquidity periods, earnings announcements, regulatory events, geopolitical shocks, macroeconomic shocks, or sudden market regime changes.
Confidence Scores
Kerdos AI may display confidence scores or similar indicators. These are model-estimated confidence indicators within the selected methodology.
Confidence scores are not:
- probabilities of profit;
- guarantees of correctness;
- guarantees of future performance;
- measures of suitability for any individual user;
- measures of how much money a user may gain or lose;
- substitutes for independent research or professional advice.
A high confidence score can still be wrong. A low confidence score does not mean that an instrument cannot perform well. Confidence indicators should be interpreted together with the methodology notes, data freshness, assumptions, risk factors, and market context.
AI-Generated Explanations
Kerdos AI may use AI to generate summaries, explanations, narratives, risk notes, or research context.
AI-generated text can be wrong, incomplete, outdated, inconsistent, or misleading. It may omit important information, overstate relationships in the data, or present uncertain interpretations in a confident tone.
AI-generated explanations should be reviewed critically and should not be treated as complete, objective, or suitable for any individual user.
Backtesting and Historical Analysis
Kerdos AI may use historical data, backtests, validation metrics, or past model behaviour to develop or evaluate models.
Backtests and historical analyses have important limitations:
- past performance does not guarantee future performance;
- historical conditions may not repeat;
- backtests may be affected by overfitting, survivorship bias, look-ahead bias, data-mining bias, transaction-cost assumptions, liquidity assumptions, and changing market regimes;
- live results can differ materially from backtested or simulated results.
Any backtest, historical score, or validation metric should be treated as research context only.
Known Limitations
Kerdos AI outputs may be affected by:
- data-quality issues;
- delayed, missing, incomplete, or inaccurate data;
- limited source coverage;
- model bias;
- overfitting;
- survivorship bias;
- look-ahead bias;
- market regime changes;
- unusual market events;
- liquidity constraints;
- corporate actions;
- currency effects;
- sector concentration;
- incorrect assumptions;
- AI hallucination risk;
- model drift;
- software errors;
- third-party provider limitations.
No Guarantee of Performance
No signal, ranking, forecast, scenario, confidence score, regime classification, model, summary, or explanation can guarantee returns or future outcomes.
Kerdos AI should not be relied on as the sole basis for investment decisions. Users remain responsible for their own research, decisions, and investment outcomes.
Output Metadata
Where available, Kerdos AI aims to display relevant context for generated outputs, such as:
- generated-at timestamp;
- data-as-of timestamp;
- price-as-of timestamp;
- forecast horizon;
- source categories;
- methodology summary;
- model version;
- assumptions;
- limitations;
- risk factors;
- conflict-of-interest disclosure;
- AI-generated content notice.
If any of this information is not available for a specific output, the absence of such information should be considered a limitation of that output.
Updates to Methodology
Kerdos AI may update its models, features, methodology, data providers, ranking logic, signal categories, forecast horizons, and AI-generated explanation process over time.
As a result, outputs generated at different times may not be directly comparable. Previous outputs may be changed, replaced, withdrawn, or no longer supported.