Urgent.News

What's breaking now, across thousands of outlets.

AI

What Solana Staking Delegation Teaches Us About AI Agent Authority

Solana's stake authority, SPL Token delegate, and liquid staking pool grant different powers. Here's how to pick the right one for an AI agent.

What Solana Staking Delegation Teaches Us About AI Agent Authority

Solana offers three options for delegating funds: stake authority, SPL token delegate, and liquid staking pool. Each delegation type grants distinct powers and comes with its own risks.

A stake authority can delegate a stake account to a validator but cannot withdraw the SOL. It can deactivate, split, merge the stake, and assign a new stake authority. The stake authority key can be compromised, leading to lost rewards and time. The withdraw authority can withdraw undelegated SOL and reset the stake authority, acting as a last resort.

An SPL token delegate assigns one address to transfer tokens up to a set cap. The delegate cannot move funds to any address, but the cap limits potential damage if compromised. If the delegate key is stolen, the attacker can only take the remaining allowance.

Liquid staking pools, like Jito and Marinade, custody SOL and issue a token representing the stake. The holder receives a claim token and no longer controls the underlying stake account's authorities. Exiting a pool involves redeeming or selling the token, which depends on liquidity and the pool's continued operation.

Despite sharing some common properties—owner's last resort authority, fixed delegate powers, and bounded damage from compromised keys—the models differ significantly in risk. Stake authorities cost rewards, SPL delegates cost remaining allowances, while liquid staking pools' upgrade authority can access all deposits. Understanding these differences is crucial when deciding which delegation model an AI agent, paying for its own compute, should utilize.

Written by urgent.news from HackerNoon's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at hackernoon.com →

More in AI

More from Thursday 29 October →