Risk Disclosure Statement

1. General Financial Risks

1.1 Risk of Loss

Prediction market trading is speculative and involves a high degree of risk. You can lose some or all of your invested capital. There is no guarantee of profit, and past performance of any strategy, agent, or simulation does not predict, guarantee, or indicate future results.

1.2 Market Risk

Prediction market prices can move rapidly and adversely based on news events, public sentiment, regulatory actions, and other factors that may be unpredictable. Markets may become illiquid, making it difficult or impossible to exit positions at desired prices.

1.3 Liquidity Risk

Prediction markets may have thin liquidity, particularly for less popular contracts. This can result in: wide bid-ask spreads; partial fills or unfilled orders; significant slippage between expected and actual execution prices; and inability to exit positions during periods of market stress.

1.4 Regulatory Risk

Prediction markets exist in an evolving regulatory landscape. The Commodity Futures Trading Commission (CFTC), Securities and Exchange Commission (SEC), state gambling commissions, and international regulators may take actions that could: restrict or prohibit prediction market trading in your jurisdiction; cause exchanges to suspend operations or delist contracts; invalidate outstanding contracts; or require reporting obligations that affect your trading activity. Sylum has no control over regulatory developments and bears no responsibility for their impact on your trading.

1.5 Exchange Risk

YorN connects to third-party prediction market exchanges (Kalshi, Polymarket). These exchanges may experience: outages, downtime, or scheduled maintenance that prevents trade execution; API changes that temporarily disable connectivity; security breaches that compromise user funds; insolvency or cessation of operations; or policy changes that affect contract settlement. Sylum is not responsible for any losses resulting from exchange-related events.

2. AI Agent and Autonomous Trading Risks

2.1 AI Model Error Risk

AI systems, including large language models (LLMs) and machine learning models used by YorN, are probabilistic and can produce incorrect, misleading, or harmful outputs. Specific risks include:

• Models may misinterpret market data, news articles, or other inputs
• Models may generate strategy code with logical errors, edge-case failures, or unintended behavior
• Models may "hallucinate" — producing confident but factually incorrect analysis or recommendations
• Model performance may degrade over time as market conditions change (concept drift)
• Models may exhibit biases present in their training data that lead to systematically poor decisions

2.2 Self-Learning and Mutation Risk

YorN's self-learning agents autonomously modify their own strategies through a mutation pipeline. This introduces compounding risks:

• Agents may modify their strategies in ways that increase risk exposure beyond your original intent
• Sequential mutations can compound, producing behavior significantly different from the original strategy
• Paper testing of mutations uses simulated conditions that may not reflect live market behavior
• The governance system (circuit breakers, harness cases) provides safety guardrails but cannot prevent all adverse outcomes
• Agent playbook lessons and evolution memory may cause future decisions to consider prior behavior that later becomes stale, incomplete, or wrong
• Decision rationale and audit evidence can explain why learned behavior influenced a proposal, but they do not make that behavior correct or profitable
• In auto-mode, mutations are applied without your manual review, increasing the risk of unintended changes
• Even in manual mode, the complexity of strategy code changes may be difficult to fully evaluate

2.3 Autonomous Execution Risk

Once deployed, agents operate continuously and may execute trades at any time, including:

• During periods when you are not actively monitoring the platform
• During periods of extreme market volatility
• At prices that differ significantly from current displayed prices (stale data risk)
• In response to events that the AI interprets differently than you would
• In coordination with other agents in team modes, where collective decisions may differ from individual agent behavior

2.4 Custom Code Execution Risk

The Strategy Lab feature generates and executes custom Python code on our servers. Risks include:

• AI-generated code may contain logical errors that produce incorrect trading signals
• Code behavior may differ between backtesting and live execution environments
• Edge cases in custom code may not be caught by validation checks
• Interaction between custom code and the execution engine may produce unexpected results

3. Backtesting and Simulation Limitations

3.1 Backtesting Limitations

• Hindsight bias: Backtests are designed with knowledge of what happened in the past
• Curve fitting: Strategies may be optimized to fit historical data but fail on new data
• Fill model assumptions: Backtests use simplified fill models (instant, probabilistic, queue position, market impact) that may not reflect actual execution conditions
• Survivorship bias: Historical data may exclude delisted or expired contracts
• Look-ahead bias: Backtests may inadvertently use future information not available at the time of the simulated trade
• Data quality: Backtests use synthetic or historical data that may contain gaps, errors, or inconsistencies

3.2 Monte Carlo Simulation Limitations

Monte Carlo simulations model thousands of hypothetical scenarios but:

• Are based on assumed return distributions that may not match actual market behavior
• Cannot predict black swan events or regime changes
• Use path-dependent models that may oversimplify real-world compounding effects
• Confidence intervals and probability estimates are approximations, not guarantees

3.3 Crowd Simulation Limitations

Crowd simulations with up to 10,000 virtual participants model market dynamics but:

• Use behavioral archetypes that are simplifications of real trader behavior
• Social influence graphs (Watts-Strogatz) are stylized models, not representations of actual market participant networks
• LLM-generated interpretations of simulation outputs are opinions, not predictions
• Simulation results predict crowd behavior, not actual event outcomes

4. Technology and Infrastructure Risks

4.1 System Failures

Technology failures may prevent trade execution, cause incorrect execution, or result in data loss:

• Server outages, network disruptions, or hardware failures
• Software bugs, including in the execution engine, ML pipeline, or agent runtime
• API rate limiting or throttling by exchanges or LLM providers
• WebSocket disconnections that delay or prevent real-time data updates
• Database failures that could affect trade records or agent configurations

4.2 Cybersecurity Risk

Despite our security measures, the platform may be vulnerable to:

• Unauthorized access to your account or exchange API keys
• Distributed denial-of-service (DDoS) attacks that disrupt the Service
• Supply chain attacks through third-party dependencies
• Social engineering attacks targeting users or infrastructure

4.3 Data Accuracy Risk

Market data, prices, and other information displayed on the Service are sourced from third parties and may be delayed, inaccurate, or incomplete. Trading decisions based on inaccurate data may result in financial loss.

5. Third-Party and Counterparty Risks

• Exchange counterparty risk: Your funds held on Kalshi, Polymarket, or other exchanges are subject to the financial health and operational integrity of those exchanges
• LLM provider risk: Changes to LLM provider APIs, pricing, or policies may affect the availability or quality of AI-powered features
• Marketplace strategy risk: Strategies shared by other users on the marketplace are not vetted, audited, or guaranteed by Sylum
• Payment processor risk: Subscription payments are processed by Stripe; disputes or processing failures are subject to Stripe's policies

6. Jurisdictional and Legal Risks

• Prediction market trading may not be legal in all jurisdictions. You are solely responsible for determining the legality of your use of the Service in your jurisdiction.
• Tax treatment of prediction market gains and losses varies by jurisdiction. You are responsible for complying with all applicable tax laws and reporting requirements.
• Changes in law or regulation may affect your ability to use the Service, access your funds, or settle contracts.
• Sylum is not licensed as a broker-dealer, investment advisor, or financial institution in any jurisdiction and does not provide regulated financial services.

7. Your Responsibilities

By using YorN, you acknowledge that you have read and understood this Risk Disclosure Statement, and you agree that:

• You have sufficient knowledge and experience to evaluate the risks of prediction market trading and AI-assisted autonomous trading
• You will only trade with funds you can afford to lose entirely
• You will actively monitor your agents and their performance, particularly after deploying new strategies or applying mutations
• You will use paper trading mode to test strategies before deploying with real funds
• You will set appropriate risk limits (position sizes, stop losses, circuit breaker thresholds) for your agents
• You will seek independent professional financial, legal, and tax advice if you are uncertain about any aspect of prediction market trading
• You understand that learned behavior is not guaranteed, can be wrong, remains your responsibility to monitor, and is not financial advice
• You assume all risks associated with the use of the Service, including but not limited to the risks described in this document

8. Contact Information

If you have questions about this Risk Disclosure Statement, please contact us: