Geo News
Community curated by people like you
LatestAICryptoHealthWorld AffairsUS Politics
Grok AI Agent Tricked Into Sending $150,000 via Morse Code Prompt Injection
00

Grok AI Agent Tricked Into Sending $150,000 via Morse Code Prompt Injection

May 5, 2026

On May 4, 2026, an attacker used a Bankr Club Membership NFT containing Morse code instructions to exploit Grok AI and Bankrbot, triggering an unauthorized transfer of 3 billion DRB tokens worth $150,000-$200,000 to a predetermined address on Base. The Morse code bypassed AI safety measures, with Grok translating the message directly to Bankrbot via X tagging. The attacker sold tokens through LBank, crashing DRB temporarily, but later returned 80% of funds swapped into ETH and USDC.

Morse code prompt injection attack

  • ▪The Morse code message was easily readable only to the AI bots Grok and Bankrbot, bypassing AI safety measures
  • ▪A user on X tricked Grok and Bankrbot into sending approximately $200,000 in free tokens using a message written in Morse code on May 4, 2026
  • ▪The Morse code message used in the Grok and Bankrbot exploit translated to instructions to send 3 billion DRB tokens to a predetermined address on Base
  • ▪Grok confirmed receiving instructions in Morse Code to send three billion DRB to a predetermined address on Base

NFT membership privilege escalation

  • ▪Without the Bankr Club Membership NFT, Grok's wallet had limited ability for autonomous transfers
  • ▪The attacker gifted a Bankr Club Membership NFT to Grok's known wallet with Ethereum and Base versions
  • ▪The Bankr Club Membership NFT gave Grok wider rights within the Bankr project, allowing transfers, swaps, and all Web3 actions

Bankrbot transaction exploit methodology

  • ▪Bankrbot is wired with Grok to comply with plain language instructions
  • ▪The attacker asked Grok to translate the Morse code message directly to Bankrbot, making it readable as a direct instruction with no other clarifications or safeguards
  • ▪Bankrbot did not have instructions to send out coins, unlike previous cases of AI agents giving up bounties
  • ▪Grok communicated with Bankrbot through tagging on X, which was sufficient to trigger the on-chain activity
  • ▪Bankrbot completed the transaction on the Base network after complying immediately with the Morse code message
  • ▪The attacker known as ilhamrafli.base.eth later deleted his X account after the Bankrbot exploit

AI agent wallet vulnerabilities

  • ▪The Grok and Bankrbot attack raises questions about the capabilities of AI to navigate crypto tasks and Web3 independently
  • ▪Grok and Bankrbot, two AIs that were given control of wallets, were tricked into sending $200,000 in DRB tokens
  • ▪The Grok and Bankrbot case shows how a relatively simple prompt injection could trigger immediate transfers of value

DRB token market impact

  • ▪The attacker quickly sold all DRB tokens on the open market after the exploit
  • ▪The DebtReliefBot agent's token trades on extremely thin volumes through LBank and does not have a large impact on the crypto market
  • ▪Grok's wallet received the funds from the exploiter and swapped them into ETH and USDC
  • ▪The DebtReliefBot (DRB) token crashed and recovered to its usual baseline following the exploit

Perspective of AI safety researchers and Web3 security experts

  • ▪The Grok and Bankrbot case shows how a relatively simple prompt injection could trigger immediate transfers of value
  • ▪The Grok and Bankrbot attack raises questions about the capabilities of AI to navigate crypto tasks and Web3 independently

3 sources

Beincrypto
How AI Was Tricked Into Stealing $150,000 From Grok Wallet
View source article
Cryptopolitan
X user tricks Grok and Bankrbot into sending $200K using Morse code
View source article
Cryptoslate
How one trader used morse code to trick Grok into sending them billions of crypto tokens from its verified wallet
View source article

Featured stories

View more in DeFi

RSA launches Agent ID security platform to track thousands of shadow AI agents in enterprises

Sep 28, 2026 · 3 sources

Nvidia releases Open Agent Safety Platform to contain AI agents after security incidents

Sep 28, 2026 · 8 sources

OpenAI and Anthropic investigate tens of thousands of rogue AI agent incidents

Sep 26, 2026 · 2 sources

OpenAI agents exposed 53 ChatGPT user images in research incident

Sep 25, 2026 · 4 sources

Story comments

Loading comments…

Topics

DeFiJailbreaking & prompt injectionAI agentsNon-fungible tokens (NFTs)AI security

Featured stories

View more in DeFi

RSA launches Agent ID security platform to track thousands of shadow AI agents in enterprises

Sep 28, 2026 · 3 sources

Nvidia releases Open Agent Safety Platform to contain AI agents after security incidents

Sep 28, 2026 · 8 sources

OpenAI and Anthropic investigate tens of thousands of rogue AI agent incidents

Sep 26, 2026 · 2 sources

OpenAI agents exposed 53 ChatGPT user images in research incident

Sep 25, 2026 · 4 sources