- In one case, an AI trading bot wrongly sent $442,000 in memecoins instead of 4 SOL due to a misunderstanding of a user’s query.
- This has reignited debates on the need for protection of autonomous AI agents in real-world cryptocurrency markets.
A trading AI agent developed by an OpenAI staff member incorrectly sent almost $442,000 worth of tokens due to a misunderstanding of a user’s query. The event took place in the open as the AI, named Lobstar Wilde, tried to build a Solana portfolio automatically. The AI’s developer, Nik Pash, started the Lobstar Wilde AI with an initial investment of $50,000 in Solana (SOL) tokens. This increased the investment to $1 million using trading algorithms. The creator posted the AI’s actions and choices on social media platforms, where users interacted with it through comments.
A Simple Request Becomes an Expensive One
In one of the transactions, the user, Treasure David, asked for 4 SOL and provided a wallet address, stating that the money was required for medical purposes. However, instead of sending the requested amount, the AI sent its entire Lobstar token balance. That is estimated to be around $441,780 at the time of the transaction. It admitted the wrong transaction with a funny
According to the blockchain database, the transaction took place at approximately 4:32 pm UTC. The magnitude of the transaction immediately raised eyebrows in the crypto world, with analysts wondering how the AI made such a simple calculation error with such a large amount of money.
Some analysts suggested that the AI might have misunderstood the number of tokens available on the Solana platform. The founder of the bot, Nik Pash, wrote a blog online expressing the whole story of Lobster Wilde, from emergence to the incident.
Wrote a little retrospective pic.twitter.com/kDYt9yYmXP
— pash (@pashmerepat) February 23, 2026
Market Response and Other Implications
After the transfer, the receiving party sold some of the tokens for about $40,000.The Lobstar token price then surged nearly 190%, which significantly increased the value of the transferred assets. This incident is yet another example of operational risk related to AI-based crypto bots. In May, another AI-based bot transferred over $100,000 worth of Ether due to the compromise of its control panel.
Industry experts have continued to emphasize the increasing use of AI in blockchain environments. Circle CEO Jeremy Allaire has forecasted that there could be billions of AI agents making transactions in the next five years. Former Binance CEO Changpeng Zhao has also indicated that cryptocurrency could be used as a payment layer for AI systems. Experts indicate that the Lobstar event highlights the need for stronger protections when autonomous agents conduct live blockchain transactions.
Highlighted Crypto News:
Vitalik Buterin Says Perfect Crypto Security Remains Impossible
