Virtuals Protocol Shields Agent Wallets from Malicious Prompt Injection

Virtuals Protocol has introduced programmable agent wallets to protect against evolving prompt injection threats. These wallets are designed to prevent unauthorized fund transfers even if an attacker successfully poisons an agent's memory or injects malicious prompts.

The new wallets operate on Base and will expand into Solana, giving agents non-custodial multi-chain smart accounts. They come with presets like 'DENY_ALL,' which requires manual approval for every transaction, and 'ACP_ONLY,' which limits where funds can actually go.

Policies are enforced server-side, independent of the AI agent's logic. This means that even if an attacker successfully compromises the agent's behavior, the wallet itself won't execute transactions that violate the owner's rules.