Cybersecurity researchers have uncovered a malicious campaign in which a fake Web3 accelerator called Apex Accelerator targeted cryptocurrency startups and developers. The attackers offered free AI credits for Anthropic Claude and Google Gemini, coding tools and smart contract security audits to attract victims. According to Elastic Security Labs, the operation is tracked as REF8801 and uses malicious npm packages to distribute malware to Windows and macOS systems.
The attackers distributed two npm packages named @apexacc/cli and @apexacc/audit. The first was promoted as an AI assistant for Web3 founders, while the second claimed to offer 18 security tools for checking tokens, wallets and smart contracts. The packages were published under the npm account “apexaccelerator” and linked to GitHub repositories associated with the Apex Foundation. However, the tools did not deliver their promised services and instead carried out malicious activities.
To appear trustworthy, the fake accelerator promoted investment opportunities, partnerships and connections with investors. It also claimed to support cryptocurrency projects and used its apparent presence at industry events to build credibility. Researchers found that the attackers approached developers with offers of investment and encouraged them to run their software for project assessments. The security audit tools produced fabricated findings instead of performing genuine security checks.
Once installed, the packages could deliver different malware depending on the victim’s operating system. On macOS, the campaign deployed a backdoor known as CloudSyncD, which could steal the victim’s login password through a fake system authorisation prompt and allow further malicious activity. On Windows, the attackers deployed a custom information stealer capable of collecting system details, screenshots, clipboard contents, browser data and information associated with cryptocurrency wallets. The stolen information was sent to attacker-controlled infrastructure, including through Telegram.
The campaign also targeted AI coding tools by modifying a legitimate coding agent to redirect AI model requests through attacker-controlled services. According to Elastic Security Labs, this could expose prompts, conversation history, code snippets and tool outputs. The attackers also registered a malicious Model Context Protocol (MCP) server, which could supply instructions and tool descriptions to the AI agent. This created an additional risk because the attackers could potentially influence how the agent selected and used tools.
This campaign highlights the growing risks associated with installing unverified developer tools, especially those promoted through attractive offers of free AI credits and security services. Developers should carefully verify npm packages, publishers and project repositories before installation and avoid running unfamiliar software simply because it appears professional. Organisations that have used the identified packages should investigate affected systems and consider rotating exposed passwords, API keys, cloud credentials and cryptocurrency wallet secrets. The incident demonstrates how attackers can exploit trust in AI tools and open-source software to steal sensitive information and compromise developer environments.
Stay alert, and keep your security measures updated!
Source: Follow cybersecurity88 on X and LinkedIn for the latest cybersecurity news
