Bybit launched AI Sub-Accounts on May 20, forcing trading bots into isolated accounts with configurable leverage and position size limits. Hard caps protect the main portfolio from a single bot failure.
Alpha Score of 28 reflects poor overall profile with weak momentum, poor value, weak quality. Based on 3 of 4 signals – score is capped at 90 until remaining data ingests.
On May 20, Bybit rolled out AI Sub-Accounts, a framework that forces every external AI agent and third-party trading tool into a segregated sub-account with hard limits on leverage and position sizes. A single bot malfunction can no longer liquidate a user's main portfolio.
The feature integrates directly into Bybit's existing AI Hub, which provides infrastructure for connecting large language model agents and automated trading bots. The sub-account layer adds risk management that was previously absent from the equation.
The core mechanism is simple: a trader creates a dedicated sub-account, allocates capital to it, and connects an AI agent solely to that sub-account. The agent cannot access the user's primary Bybit account. Its API key is scoped to the sub-account's permissions.
For a retail user experimenting with a trading bot for the first time, the benefit is immediate. You allocate $500 to a sub-account, let the AI agent trade within that box, and know that your other holdings are completely untouched regardless of what the bot does. The sub-account acts as a sandbox with real capital and a fixed ceiling.
The real value lies in the leverage and position size limits being enforced by the exchange, not by the bot's code. Users can configure maximum leverage ratios, cap position sizes, and define exactly what each AI agent is allowed to do through granular API permissions. If the bot hits its limits, it stops. The exchange rejects the order at the API gateway.
Key insight: Hard caps at the account level are a reliable circuit breaker. They do not require the bot to be well-written. They only require the exchange to honor its own API settings.
A common error among traders using automated tools is assuming that the bot's built-in risk controls will work under all conditions. In reality, AI agents, especially those running large language models or connecting through third-party infrastructure, can behave unpredictably when market conditions change rapidly or when their data feed lags.
Consider an AI agent programmed to trade Bitcoin (BTC) futures with a maximum of 5x leverage and a position size cap of 1 BTC. Without sub-account isolation, a single coding error or a corrupted input could cause the agent to submit an order for 50x leverage on 10 BTC. If the exchange accepts that order, the position could be liquidated instantly, wiping out the entire account balance.
With Bybit's AI Sub-Accounts, that order would be rejected at the API gateway because it exceeds the preconfigured limits. The main account remains untouched. The trader simply sees an error message instead of a liquidation.
Restricting API permissions to the sub-account level also reduces the attack surface. If a third-party bot's infrastructure is compromised, the hacker can only access the capital in that specific sub-account. The main account's funds and positions are unreachable.
For traders running multiple bots, each with its own strategy, this compartmentalisation is a significant security upgrade. The configurable limits include:
The sub-account is not a portfolio-wide circuit breaker; it is a per-bot firewall. Manual trading on the main account while a bot is active still exposes the main account.
For professional traders, the sub-account framework solves a coordination problem. Running multiple AI agents in parallel, each with different strategies and risk profiles, previously required either a single account with complex order routing or multiple exchange accounts with different login credentials.
Bybit's approach lets each agent operate in its own sandbox with individually defined risk parameters. If one strategy blows up, the others continue normally. The sub-accounts are all managed through Bybit's existing AI Hub, which provides monitoring and management tools for the agents themselves.
Setting the caps too tight can strangle a strategy that needs flexibility. Setting them too loose defeats the purpose of the sub-account. Bybit allows configurable limits, so a user can calibrate each sub-account to the specific strategy's needs.
Practical rule: Allocate capital to each sub-account based on the maximum acceptable loss for that strategy, not the expected return. The sub-account limits become the hard risk budget. The bot's internal logic manages the more granular decisions.
Bybit's move is a response to the growing popularity of AI trading agents in crypto markets. Other exchanges may follow with similar features. For now, Bybit is the first major platform to offer account-level isolation specifically for automated bots.
Traders testing the feature should start with a small allocation and verify that over-limit orders are rejected immediately. The true test will come during a flash crash or fast market, when every millisecond of API latency matters. A sub-account that takes 500 milliseconds to reject a rogue order might still allow a partial fill before the rejection takes effect.
For more context on broader crypto market infrastructure, see our crypto market analysis and Bitcoin (BTC) profile. For a related discussion on regulatory guardrails, read EU Rewrites MiCA Stablecoin and DeFi Rules Before July Deadline.
Prepared with AlphaScala research tooling and grounded in primary market data: live prices, fundamentals, SEC filings, hedge-fund holdings, and insider activity. Each story is checked against AlphaScala publishing rules before release. Educational coverage, not personalized advice.