Fhenix Upgrades Sedona With Confidential FHE to Keep Balances Hidden From AI Agents
Self-custodial trading platform Sedona has partnered with Fhenix to integrate confidential fully homomorphic encryption (FHE) infrastructure into its Arbitrum-based system, the companies announced on July 15.
The upgrade replaces Sedona’s existing security model based on trusted execution environments (TEEs) with fully homomorphic encryption, which allows computations to be performed on encrypted data without revealing it. The change aims to keep user balances, portfolio positions, and artificial intelligence agent spending limits encrypted even during processing, reducing reliance on trusted hardware or operator-based models.
“Sedona is exactly the kind of application Confidential FHE was built for,” said Guy Itzhaki, CEO of Fhenix. “Trading platforms and financial applications need privacy that extends beyond transactions to balances, positions and increasingly the parameters that autonomous agents operate within. By moving from trusted hardware to cryptographic guarantees, Sedona is showing how confidential finance can become a native capability on Arbitrum rather than an optional feature.”
Sedona, founded by Tyler Maxwell, offers spot trading, perpetuals, and structured products. The platform is migrating from the Seismic ecosystem to Arbitrum, and the FHE deployment will follow completion of that migration, shifting the privacy model from hardware-based trust assumptions to cryptographic guarantees.
In a statement, the companies noted that many private DeFi solutions rely on TEEs, which require users to trust the underlying hardware, or committee-based approaches requiring trust in a set of operators. FHE eliminates these trust assumptions by enabling computations on encrypted data without decryption.
“We started with TEEs because they were the most practical way to deliver privacy, but our goal has always been to remove trust assumptions wherever possible,” Maxwell said. “Fully homomorphic encryption lets us protect sensitive financial data through mathematics rather than hardware, providing a much stronger foundation for the future of self-custody.”
Under the integrated model, Sedona users’ balances and positions will be hidden from AI agents executing inside the system as well as from other parties. The technical integration is ongoing; specific timing for the rollout on Arbitrum was not provided beyond that it will follow completion of the migration.