The BNB Chain network announced today, May 16, the incorporation of Bankr into its ecosystem, an artificial intelligence (AI) agent capable of executing cryptocurrency transactions autonomously, which allows paying for access to more than 30 AI models with stablecoins from BNB Chain wallets. The transaction fees They are subsidized by Bankraccording to the BNB Chain announcement, so users will not bear that additional cost when paying for the models. According to Bankr, the integration allows access to models such as Claude, GPT, Gemini, Grok, Deepseek, Qwen, Kimi and Minimax from a single connection pointinstead of contracting and configuring each model separately with its own provider and access key. This means that users do not need to move funds between chains or manage individual access keys for each model. Payment is made in USDT or USDC from the user’s wallet with automatic recharging of the balance when it falls below a defined threshold, which prevents a work session from being interrupted due to lack of funds in the middle of a query, as explained by the team behind Bankr.
Welcome @bankrbot to the BNB ecosystem!You can now fund your agents on 30+ AI models (Claude, GPT, Gemini and more) with stablecoins on BSC directly, without bridging or extra gas.Note: This post is for informational purposes only and not financial advice. DYOR. https://t.co/IoqCpWstaX—BNB Chain (@BNBCHAIN) May 16, 2026
The background that the advertisement does not mention
Bankr is also the platform that starred in March a USD 200,000 exploit against a wallet associated with the Grok chatbot of X on the Base network, as reported by BitcoinDynamic. In that case, an attacker manipulated the Bankr agent into executing an unauthorized transfer through an instruction disguised in Morse code– Posted a message to X that Grok translated without recognizing the embedded order, and Bankr executed it upon detecting it in the public feed. The attacker returned 80% of the funds; the remaining 20% was not recovered. The BNB Chain integration announcement does not include any statement about design changes implemented after that incident. Developer Vadim, who analyzed the exploit, pointed out that the structural problem is not the language model but the authorization architecture, since as long as Bankr uses text generated by an AI as an instruction to execute financial transactions, any manipulation of that text can result in an unauthorized transfer.
