Sender AI — agentic threat model

Not certain from the listing — The specific foundation models powering Sender AI's intent-centric translation are not disclosed. Standard risks include adversarial prompt injection that could trick the model into generating malicious transaction payloads.

Not certain from the listing — The data operations, vector stores, and RAG pipelines used to parse user intent and blockchain state are unspecified. Risks include data poisoning of the knowledge base used to interpret smart contract ABIs.

Sender AI's core framework translates user intentions into automated, on-chain operations. The primary threat here is tool misuse and insecure tool integration, where a compromised planning loop or hijacked intent parser executes unauthorized smart contract calls or transfers assets to malicious addresses.

Built on the OP Stack and supported by NEAR Protocol, the deployment infrastructure relies heavily on layer-2 scaling solutions and decentralized network nodes. Threats include smart contract vulnerabilities in the execution network, consensus-level attacks, or compromised RPC endpoints.

Not certain from the listing — There is no mention of real-time transaction monitoring, anomaly detection, or guardrails to intercept malicious or erroneous AI-generated transactions before they are committed to the blockchain.

Not certain from the listing — While the project is open-source, specific compliance frameworks, key management policies (such as MPC or non-custodial wallet security), and formal audits are not detailed in the directory listing.

Sender AI establishes a decentralized transaction execution network utilizing 'smart agents'. This multi-agent ecosystem is highly vulnerable to agent-to-agent trust abuse, where rogue or compromised agents propagate malicious transaction intents or exploit cascading failures across the decentralized network.