Sui Introduces Agent Pass to Cap Losses From Leaked AI Agent Keys at User-Set Limits
Key Takeaways
- •Sui introduced Agent Pass at the Sui Basecamp ecosystem conference as a permission framework for AI agents operating on its layer-1 network.
- •The framework restricts an AI agent's permitted actions to a scope that the account holder defines in advance.
- •If an agent's cryptographic key is compromised, Agent Pass limits the resulting losses to a ceiling chosen by the account holder.
- •The design aims to prevent a leaked key from granting unlimited access to all assets the credential can reach, replacing unbounded exposure with owner-set limits.
- •The announcement did not specify how the loss caps are technically enforced or when the framework will become broadly available.
Sui Introduces Agent Pass to Cap Losses From Leaked AI Agent Keys at User-Set Limits
October 15, 2026 — COINOTAG
Sui has introduced Sui Agent Pass, a permission framework that keeps an AI agent's actions within a range the user defines and, if the agent's key is leaked, caps the resulting losses at a ceiling set by the account holder. The feature was presented at Sui Basecamp, the Sui ecosystem conference.
The announcement addresses a well-known risk in on-chain automation: AI agents typically execute tasks by signing transactions with cryptographic keys, and if those credentials are exposed, any assets the key can access may be at risk. Because whoever holds the key can sign as the agent, the credential itself becomes the attack surface once it leaves the owner's control. Agent Pass is designed to contain that exposure in advance, by letting the account holder specify ahead of time how far an agent's authority extends and how much can be lost in the event of a compromise.
The design changes the worst case for users of automated accounts. Rather than a leak opening unbounded access to everything the key can reach, the potential damage is bounded by limits the owner set beforehand — an approach whose weight grows as more on-chain activity is delegated to autonomous agents instead of being signed manually, transaction by transaction.
With user-set limits on both permitted actions and potential losses, Sui positions the framework as a safeguard for users deploying autonomous AI agents on the layer-1 network.
The announcement did not detail how the caps are technically enforced or when the framework will be broadly available — specifics that will clarify how the model works in practice.
This content was first published on COINOTAG: https://en.coinotag.com/sui-agent-pass-caps-losses-leaked-ai-agent-keys
