Sedona, a self-custodial trading platform and neo-bank, has teamed up with Fhenix to bring confidential fully homomorphic encryption (FHE) to its Arbitrum-based system. The upgrade is designed to keep sensitive financial data — including user balances — hidden from AI agents that increasingly operate on decentralized networks.
Why AI agents need to be kept in the dark
AI agents are becoming common in decentralized finance, executing trades, managing portfolios, and interacting with smart contracts. But these agents can also scrape on-chain data, exposing users' financial positions. Sedona's integration of FHE aims to prevent that. By encrypting data at the protocol level, the platform ensures that even when an AI agent processes a transaction, it never sees the underlying balance or other private information.
How FHE works in this setup
Fully homomorphic encryption allows computations to be performed on encrypted data without ever decrypting it. Fhenix provides the confidential FHE infrastructure that Sedona is plugging into its Arbitrum-based architecture. That means balances, trade histories, and other key financial details stay encrypted throughout any AI-driven operation. Users retain full control of their data — a core promise of self-custodial platforms.
Self-custody meets privacy
Sedona already lets users hold their own private keys, a model that cuts out third-party custodians. The FHE layer adds another privacy wall. Even if an AI agent is granted permission to execute a trade on a user's behalf, it won't be able to read the user's total balance or past activity. The combination of self-custody and encryption could appeal to traders who want automated strategies without sacrificing privacy.
The partnership is still in its early stages. Neither company has announced a specific rollout date for the FHE upgrade. But the move signals a growing recognition that as AI agents proliferate in DeFi, the tools to keep financial data hidden need to keep pace.
