Automated and AI-Assisted Trading Disclosure

This disclosure applies to automated or AI-assisted activity on an ORBITRA ONE™ account: Cortex agents, strategies built and deployed with Strategy Forge, algorithmic order types, copy portfolios, bots licensed through the marketplace and programs connected through APIs.

It supplements the general risk disclosure and the trading and leverage risk disclosure.

Cortex follows a governed loop. An agent observes market and portfolio data, reasons with an ensemble of models, proposes a trade intent, simulates its effect with Aegis and checks it against your policy. Only an authorized intent is sent to ApexMatch for execution.

Every agent and deployed strategy acts within a policy that you define and can change at any time. A policy can set:

• the maximum capital an agent or strategy may allocate;
• daily and lifetime loss limits;
• a leverage ceiling for each instrument;
• the markets it may and may not trade;
• session windows and an expiry time for its authority;
• the actions that require your confirmation;
• the data sources it may rely on;
• a one-command kill that stops activity and revokes its permissions.

Policies limit what automation may do. They do not make its decisions correct: an action that respects every limit can still lose money.

Automation, including AI-assisted automation, can fail or behave unexpectedly. The principal sources of that risk are:

• Model error — models simplify markets and can misread conditions, rely on patterns that do not persist or produce confident outputs that are wrong. AI-generated explanations can sound plausible and still be incorrect.
• Data errors — automation acts on the data it receives. Delayed, incomplete, erroneous or manipulated market data, reference prices or external inputs lead to decisions based on a false picture.
• Overfitting — a strategy tuned closely to past data can capture noise rather than a durable effect and perform poorly on new data.
• Regime change — relationships between prices, volatility, liquidity and correlation shift, sometimes abruptly, and a strategy built for one regime can fail in another.
• Latency — delays between signal, decision and execution, whether from networks, queues, risk checks or confirmation steps, can cause fills at prices different from those modeled, or no fill at all.
• Unexpected behavior — automation can interact with other automated participants, with its own orders or with market rules in ways its designer did not anticipate, including rapid or repeated order activity.
• Technical failure — software defects, outages, lost connectivity or third-party failures can stop automation, prevent cancellations or leave positions unmanaged.

Backtests, stress tests, paper trading and track records help evaluate a strategy. Past or simulated performance, however rigorously produced, does not predict future results.

• Backtests apply a strategy to past data under assumptions about costs, fills and slippage. They benefit from hindsight and cannot reproduce the market impact of real orders.
• Stress and scenario tests show behavior under chosen shocks; real events can be more severe or take a different shape.
• Paper trading simulates execution against real-time market data without committing capital, so it cannot reflect queue position, partial fills or how other participants react to your orders.
• Realized track records reflect a particular period, capital size and set of conditions. Results differ between followers because of timing, allocation size, fees and slippage.

Each Strategy Forge version stores its code hash, data window, assumptions, costs, risk limits, approvals, execution receipts and realized behavior. This trust record makes results verifiable; it does not make them repeatable.

You decide whether to use automation and on what terms. Actions taken within the permissions you grant are treated as your instructions. You are responsible for:

• granting only the permissions an agent or strategy needs, and reviewing them regularly;
• setting capital, loss, leverage, market and time limits consistent with the risk you are prepared to take;
• monitoring activity, positions and notifications, including outside your usual hours;
• using the kill command or revoking permissions promptly when behavior is not what you expect, and then reviewing any orders and positions that remain open;
• keeping API keys, agent credentials and devices secure;
• understanding the logic, leverage and fees of any strategy you deploy or follow.

You can require confirmation for classes of action — for example, orders above a set size, trades in new markets, increases in leverage or transfers. Each request shows the proposed action, the agent’s stated reason, the simulated risk and the data sources behind it.

Confirmation is a control, not a guarantee. Assess each proposal on its merits: once you approve it, the decision is yours. Requests that are not confirmed expire under your policy settings. Actions required by margin rules, such as liquidation, are not subject to confirmation.

Automated activity is logged. Each action carries a receipt linking the agent or strategy version, the policy in force, the simulated risk, the authorization decision and the resulting orders and fills.

Receipts let you review, reconcile and dispute activity. Logs are also used to investigate incidents, meet record-keeping obligations and support market surveillance.

Cortex analysis, suggestions, rankings, alerts and strategy templates are general tools. Unless a separate written agreement expressly provides otherwise, they are not personalized investment advice and do not assess whether a transaction is suitable for you, because they do not take account of your full financial situation, objectives or needs.

AI-generated explanations describe a model’s reasoning as far as it can express it, and they may be incomplete or inaccurate. Seek independent advice where appropriate.

Strategies, copy portfolios and bots in the marketplace are created by third parties unless labeled otherwise. A listing means that the strategy met publication requirements; it is not an endorsement, a recommendation or an assurance of future results.

Copying a portfolio replicates a creator’s decisions in your account using the allocation method you choose — fixed capital, volatility target, risk budget or proportional copy — subject to your exclusions and hard stops. Your results will differ from the creator’s because of timing, size, fees, slippage and your own limits.

Creators can modify, pause or withdraw a strategy. Version changes appear in the trust record, and you should review them and your allocation when they occur. Creators may earn performance, subscription or license fees, which are shown before you allocate or subscribe and which reduce your net result.

Conflicts can arise when a creator trades the same instruments as followers, when fees depend on assets allocated or on performance, when a creator is connected with ORBITRA ONE™ or with a liquidity provider, or when we earn fees from marketplace activity.

We identify conflicts and disclose them in each strategy’s record, alongside version history, fees, slippage, real fills and model lineage. Marketplace rankings draw on verified records — return profile, drawdown, risk-adjusted performance, capacity, market regime and verification status. Where a conflict cannot be adequately managed, the strategy may be refused or removed.

If automated activity on your account looks wrong, stop it with the kill command first, then contact hello@orbitraone.com. For questions about this disclosure, write to hello@orbitraone.com.