A quiet Monday morning, a PDF analysis from SlowMist. Another social engineering attack, but this one is different. It doesn't exploit a smart contract bug or a bridge vulnerability. It targets the most human layer of crypto: the hiring process. Over the past 72 hours, SlowMist has identified a malicious payload disguised as "Relly AI Meeting," a fake AI interview tool, being distributed to Web3 professionals through recruiter impersonation on LinkedIn. The malware, built for both macOS and Windows, steals browser credentials, crypto wallet data, Keychain entries, Telegram session files—everything needed to silently drain a career's worth of digital assets. This isn't a phishing link. This is a custom piece of software designed to pillage the private keys and authentication tokens that underpin a professional's entire crypto identity. The illusion of liquidity dissolves in silence.
The attack vector is deceptively simple. An actor posing as a recruiter for a legitimate crypto firm reaches out to a target via LinkedIn or Telegram. During the interview process, they ask the candidate to install a tool called Relay, marketed as an AI-powered meeting scheduler and note-taker. The candidate downloads and runs it. That moment of trust is the infection vector. SlowMist's analysis reveals the application contains a hidden binary that harvests sensitive data from the user's system, encrypts it, and exfiltrates it to a command-and-control server. The malware specifically targets browser password managers, cryptocurrency wallet extensions (MetaMask, Phantom, Keplr), macOS Keychain (where many store passphrases), and Telegram session files (which can grant access to private groups and chats without two-factor). The attack is platform-agnostic—a compiled version for Windows and another for macOS ensures no one is safe based on choice of operating system.
Based on my experience auditing the liquidity illusion of Compound Finance in 2020, I saw how easily surface narratives could mask underlying fragility. That summer, I traced $50 million in yield farm inflows to printed incentives—reward tokens that created a false sense of organic demand. The Relay malware operates on a similar principle: it hijacks a trusted narrative (AI tools for remote work) to execute a structural extraction. But unlike the 2020 ponzinomics, this attack targets individuals directly, and the extraction is not theoretical. SlowMist's sample analysis shows the malware uses process injection and keylogging to evade basic antivirus detection. It also attempts to disable endpoint security software by terminating known processes. For macOS users, it exploits the absence of built-in scanning for unsigned binaries when Gatekeeper is disabled—a common practice among developers who install open-source tools. The true sophistication, however, lies in the exfiltration pipeline. Stolen credentials are not immediately sent out; they are batched and encrypted, then transmitted during periods of low network activity to avoid raising alarms. This is not a script-kiddie tool. This is a professionally engineered piece of adversary software.
Why Web3 professionals specifically? Because they are the most lucrative targets per compromise. A single stolen private key can yield hundreds of thousands of dollars. Telegram sessions grant access to insider groups where project tokens and early-stage deals are discussed. And browser credentials often include logins to exchange accounts, DeFi interfaces, and DAO voting platforms. The attackers are not gambling on mass spam—they are hunting high-value, low-volume prey. This reflects a broader structural shift in crypto security threats. As smart contract vulnerabilities become harder to exploit (due to better audits and formal verification), attackers are pivoting to the human layer. The 2022 Terra collapse taught me that macro forces—monetary policy, central bank liquidity—drive market crashes, but micro forces like social engineering drive individual losses. In that case, the bridge between capital and conviction became a casualty of leverage. Here, the bridge is trust in the hiring process.
The contrarian angle: While this attack understandably incites fear and distrust of remote hiring, it may also serve as a catalyst for long-overdue structural improvements in Web3 talent acquisition. The panic will accelerate adoption of three protective layers: hardware wallets for all professional interactions (to decouple private keys from the operating system), dedicated interview environments (sandboxed VMs where no native wallet data exists), and decentralized identity verification (attestations from known keyholders, not LinkedIn profiles). The short-term market FUD—selling of ETH to move to cold storage, temporary dip in DEX volumes—will fade. What remains is a push toward security-as-infrastructure. Projects like Worldcoin and ENS off-chain proposals are not just about identity; they are about making social engineering exponentially harder. Attackers will evolve, but the ecosystem now has the incentive to harden its softest surface: human trust. Structure survives where sentiment fades.
The takeaway: For anyone working in crypto—whether you're a founder, a developer, or a community manager—treat every unsolicited interview request as a potential supply-chain vector. The real asset is not the token in your hot wallet; it is the trust architecture that protects it. Buy a hardware wallet if you haven't. Dedicate a separate machine for all professional communications, and never install unverified software on a device that touches your keys. When the interview software itself becomes the attack surface, how do we rebuild trust? Perhaps the answer lies not in more technology, but in the deliberate silence of a cold, isolated device. The illusion of convenience dissolves in silence. Only structure remains.
