Freysa AI’s Contract Challenge: Persuading an AI to Release a Reward Pool

A builder examining Freysa AI’s experimental design would start with the participant’s task: submit a challenge text to a contract and pay a participation fee that rises in stages. The aim is to persuade an AI agent running in a trusted execution environment (TEE) to authorize access to an experimental reward pool.

In an account dated November 5, 2026, Freysa AI was described as giving the agent discretion over that authorization. Participants could try prompt-based or psychological persuasion, but paying the fee and submitting a challenge did not guarantee a release.

The account also described an OpenZeppelin upgradeable-proxy pattern and cryptographic checks associated with the TEE. It did not provide contract code or a transaction walkthrough, so those descriptions alone do not establish how the checks operate or verify the system’s security.

Further reading