Bybit launches AI sub-accounts to ringfence trading

Bybit has launched AI Sub-Accounts, enabling users to connect artificial intelligence trading agents to its platform. The feature is now live across the exchange.

The new account type is separate from regular, custodial and Islamic sub-accounts. It is designed to keep AI-driven trading activity separate from a user's main holdings and to reduce the risks associated with automated agents operating via API access in live trading environments.

The move comes as exchanges and traders test a growing number of AI-based tools for market analysis and automated execution. That expansion has raised concerns about what happens when an AI agent is compromised, contains flawed code or behaves unexpectedly while connected to accounts holding client assets.

The new setup is intended to draw a clear boundary between human-controlled portfolios and machine-operated trading. Under the model, AI agents are confined to a designated sub-account rather than given access to a trader's full account.

Users can set limits on the assets held in the AI sub-account, as well as transfer and leverage restrictions. Withdrawals can also be turned off, and agent commands restricted according to the account holder's preferences.

Activity carried out by the AI agent remains within the sub-account, with no cross-account movement. Parent account holders retain visibility over the agent's operations through read-only oversight.

Another built-in control is that execution occurs only via API access. There is no login access or in-app switching for the AI-controlled account, a measure intended to reduce the risk of hijacking or unauthorised manual access.

The exchange is making this structure the default for any trader connecting an AI agent to Bybit. Users will not be able to run such agents directly through their primary trading account on the platform.

Risk controls

Bybit presented the launch as a response to a changing threat profile in digital asset trading. Automated systems can act at speed and scale, but they also create new points of failure when software errors, poor permission design or malicious intervention give agents wider authority than intended.

For traders, one of the main concerns is that an agent with broad permissions could trigger unauthorised transfers, forced liquidations or outsized leveraged positions. In volatile crypto markets, those errors can quickly turn into losses and may be difficult to reverse once trades are executed on exchange infrastructure.

The AI Sub-Account is meant to give traders a way to test new agents or experimental strategies in isolation before allowing wider use. That approach mirrors risk-management practices in other areas of financial technology, where firms often separate trial activity from production accounts to limit exposure.

Victor Wu, Head of AI Agent Architecture at Bybit, said the shift reflects the need for tighter controls around automated trading.

"We recognise that as agentic trading enters the mainstream, the security baseline has to evolve. No agent should have unchecked power over a trader's full portfolio. The new and refined setup prevents AI agents from controlling a trader's entire account or moving assets unpredictably," Wu said.

He described the feature as part of a broader effort to enable users to adopt AI systems without giving up control over their funds.

"Bybit's AI Sub-Account creates a security perimeter that protects assets while allowing traders to benefit from AI innovation," Wu said.

Broader context

Crypto exchanges have spent the past several years tightening controls around custody, access rights and operational resilience after a series of hacks, failures and governance breakdowns across the sector. The arrival of AI agents adds another layer to those efforts because the technology can both improve execution and introduce hard-to-predict behaviour.

Bybit is one of several large exchanges examining how account architecture should adapt as algorithmic tools become easier for retail and professional traders to use. The launch suggests platforms are moving beyond simple API key permissions toward more segmented account structures that treat AI systems as a distinct category of user access.

For users, the practical effect is likely to be a more controlled way to deploy automated strategies while keeping primary balances outside the reach of software agents. For exchanges, the challenge will be whether such controls are enough as AI trading tools become more complex and more widely used across highly leveraged crypto markets.

Any trader connecting an AI agent to Bybit will operate through an AI Sub-Account by default, providing baseline asset protection regardless of experience level or technical knowledge.