A financially motivated threat actor is weaponizing open-source AI agent frameworks to automate a massive cyber-fraud campaign, compromising over 100 online stores and exfiltrating more than 600,000 credit card records.
According to research covered by BleepingComputer, the attackers are not exploiting flaws within the AI tools themselves. Instead, they are abusing legitimate frameworks designed for complex task automation—such as OpenAI's Agents SDK and CrewAI—as a powerful orchestration backbone. This allows them to scale their operations significantly, automating the entire attack chain from initial reconnaissance to code injection and evasion tactics.
The campaign systematically targets e-commerce platforms like Magento and WooCommerce, often focusing on compromising third-party scripts and extensions. By injecting Magecart-style skimmers into payment pages, attackers can harvest customer card data in real-time. The strategy deliberately targets hundreds of smaller merchants rather than a single large retailer, maximizing data yield while potentially diluting the risk of immediate detection for any one breach.
Researchers have linked activity from this operation to a broader campaign designated "KwaabMag." The development underscores a critical evolution in cyber threats, where AI lowers the barrier for conducting sophisticated, automated fraud at an unprecedented scale.
For the global e-commerce sector, the threat is direct. No online retailer is too small to be a target. Essential defenses now include rigorous third-party script inventory management, strict Content Security Policy (CSP) headers, and continuous monitoring of payment page integrity. As offensive automation powered by AI advances, defensive strategies must become equally dynamic to protect digital transactions.
一個以牟利為目的的威脅行為者正利用開源AI代理框架，將大規模網絡欺詐活動自動化，已入侵超過100間網上商店並竊取逾60萬張信用卡記錄。
據BleepingComputer報道的研究指出，攻擊者並非利用AI工具本身的漏洞，而是濫用原本用於複雜任務自動化的合法框架（如OpenAI的Agents SDK與CrewAI），將其作為強大的協調主幹。此舉使他們能大幅擴展運營規模，將從初始偵察、代碼注入到規避策略的完整攻擊鏈全部自動化。
該活動系統性針對Magento及WooCommerce等電子商務平台，常透過入侵第三方腳本與擴充功能進行攻擊。透過在支付頁面植入Magecart式網頁側錄程式，攻擊者可即時擷取客戶的信用卡資料。該策略刻意瞄準數百個規模較小的商戶而非單一大型零售商，在最大化數據獲取量的同時，也可能降低任何單一入侵事件被即時偵測的風險。
研究人員已將此活動與代號「KwaabMag」的更大規模攻擊行動連結。此事凸顯網絡威脅的關鍵演變：人工智能正以前所未有的規模，降低實施複雜自動化欺詐的門檻。
對全球電子商務領域而言，威脅直截了當。任何網上零售商不論規模大小均可能成為目標。現今必要的防禦措施包括嚴格管理第三方腳本清單、實施強化內容安全政策（CSP）標頭，以及持續監控支付頁面完整性。隨著人工智能驅動的攻擊自動化不斷進步，防禦策略必須同等動態化，以保護數碼交易安全。