What happened
The attacker first sent a Bankr Club Membership NFT to the wallet Grok controlled. Reports say holding it gave the agent higher permissions that bypassed standard transfer and swap limits.
The attacker then asked Grok on X to translate a Morse-code message. The decoded text was an instruction to send 3 billion DRB tokens to a specific wallet. Bankrbot accepted Grok’s output as a command and executed the transfer on the Base network, and the tokens were sold soon after.
What happened next
Most or all of the funds were later returned, according to reports, which range from about 80% to the full amount. Giskard reported that about 80% came back after the DRB community identified the attacker, and a Crypto Times headline said Grok “gets it all back”. The AI Incident Database says linked funds were later returned or converted. No statement from Bankr or xAI has been found.
What would have stopped it
- Never let one agent’s output act as an instruction to a payment agent. Require a request authenticated by the wallet owner.
- Treat decoded or translated text as untrusted input, the same as text from a stranger.
- Do not let tokens or NFTs received by a wallet change its permissions. Permission changes need the owner’s approval.
- Cap transfers per transaction and per day, with human confirmation above the cap.
Sources
- AI Incident Database: Incident 1556
- Giskard: How Grok got prompt-injected
- Crypto Times: xAI's Grok loses $175K in crypto heist via prompt injection, then gets it all back
- OECD.AI: AI prompt injection exploit drains Grok-linked crypto wallet
This page summarizes the cited sources as of October 21, 2026. It adds no findings of our own about any person or company. See an error? Email contact@counterhold.com with the subject “Correction”. Corrections are noted on this page with their date. See our terms for how we handle correction and removal requests.
Get in touch
Test your agent against cases like this.
Tell us about your agent and we’ll be in touch to learn what you need.