The email landed in his inbox at 2:37 PM on a Tuesday. It was from a recruiter at what appeared to be a well-known DeFi protocol—logo, signature, even a LinkedIn profile that had been active for three years. The message was warm, professional, and offered a chance to join a team working on cross-chain interoperability. The hook was irresistible: a fully remote position, competitive token package, and an invite to a quick AI-assisted interview using a new tool called 'Relay.' The developer, let's call him Alex, clicked the link. He downloaded the app. He ran the installer. Within minutes, his entire crypto wallet was drained, his Telegram sessions hijacked, and his Keychain credentials lifted. The next day, Alex's worst fear became reality as he watched his ETH sweep to a contract he had never authorized.
This is not a hypothetical. This is the new face of Web3 social engineering—a malicious recruitment campaign unearthed by SlowMist on July 29, 2025, targeting the very people who build and secure our decentralized future. The attack is a masterclass in exploiting trust, combining the narrative of AI-driven hiring with custom cross-platform malware. But beneath the technical sophistication lies a deeper truth: Web3's trust model, which we have celebrated as permissionless and open, has a dark underbelly when applied to human relationships. We code for trustlessness, but we hire for trust. And that gap is now being weaponized.
Context: The Open Door of Web3 Hiring
The blockchain industry has always prided itself on being a meritocracy. Your wallet is your resume. Your GitHub commits are your references. Your DAO contributions are your reputation. In a world where anyone can join a protocol's Discord or contribute to a Gitcoin grant, the barrier to entry is low—and so is the barrier to exploitation. Remote work is the norm, and the hiring process often relies on a combination of asynchronous communication, video interviews, and shared documents. The rise of AI-powered tools for meeting scheduling, note-taking, and candidate screening has only accelerated this trend. Protocol teams are hungry for talent, and candidates are hungry for the next big opportunity.
But this openness creates a fertile ground for attackers. They impersonate recruiters from legitimate projects—often using stolen or forged identities from platforms like LinkedIn or Telegram. They weaponize the very tools we trust: Zoom, Notion, and now AI meeting apps. The SlowMist report details a campaign where the attacker posed as a recruiter for a respected Web3 company, invited the target to an interview using a fake 'Relay' AI meeting software, and convinced them to install a malicious installer. Once executed, the malware—coded for both macOS and Windows—stole browser credentials, cryptocurrency wallet data, system Keychain entries, and Telegram session tokens. The attacker didn't just want access to a wallet; they wanted access to the entire digital identity of the victim.
This is not a simple phishing link. This is a tailored attack that demonstrates a deep understanding of the Web3 professional's workflow. The attacker knew that the victim likely had their private keys in a hot wallet for daily operations, their Telegram logged in for team communication, and their Keychain storing passwords for exchanges. The malware was not just a keylogger; it was a comprehensive data harvester designed to maximize the surface area of exploitation.
Core: The Architecture of Betrayal
Let's dissect the technical anatomy of this attack. The malware, which SlowMist had the chance to analyze after the incident, operates cross-platform—a sign that the attacker invested significant development resources. On macOS, it leverages the system's Keychain access to extract saved passwords and cryptographic keys. On Windows, it targets the credential manager and browser-stored secrets. But the real genius is in its targeting of Telegram.
Telegram sessions are often used as the primary communication channel for Web3 teams. Once an attacker gains a valid session token, they can impersonate the victim in DMs, group chats, and channel administrations. This opens the door for a secondary wave of attacks: sending malicious links to colleagues, requesting fund transfers from DAO treasuries, or spreading the malware further under the guise of a trusted peer. In one example, the attacker used the victim's Telegram account to message their project's multisig signers, asking them to approve a 'critical upgrade' that was actually a drain transaction.
The malware also scrapes browser data—bookmarks, cookies, and autofill entries—to identify which crypto exchanges, DeFi apps, and wallet dashboards the victim uses. With this information, the attacker can perform session hijacking, bypassing 2FA for platforms where the session cookie is still valid. The result is a multi-layered theft that goes beyond a single wallet drain.
From a cryptographic perspective, the attack does not break any protocol. It does not exploit a smart contract vulnerability. It does not crack ECDSA. Instead, it exploits the human side of the trust model—a vulnerability that no formal verification can fully eliminate. The security of the system is only as strong as the weakest link in the chain of social interaction. As I often say in my writings: 'Code is law, but people are the soul.' Here, the soul was deceived, and the code silently obeyed.
The implications for the Web3 ecosystem are profound. We have spent years building trustless infrastructure—permissionless blockchain layers, non-custodial wallets, decentralized identity standards. Yet the onboarding process for new users, employees, and contributors remains stubbornly centralized and trust-heavy. When a protocol team hires a core developer, they rely on a mix of references, interviews, and gut feeling. There is no on-chain attestation of the recruiter's identity, no zero-knowledge proof of their role at the company, no verifiable credential that the job offer is genuine.
Contrarian: Don't Blame the Technology, Blame the Ritual
A natural response to this attack is to call for better technology: hardware wallets for all, isolated virtual machines for interviews, air-gapped computers for signing transactions. And yes, these measures would reduce the attack surface. But they miss the point. The core vulnerability is not in the operating system or the wallet software. It is in the ritual of trust that underpins Web3 hiring.
We have developed a culture where the first interaction with a potential employer is often a DM from a stranger, followed by a link to a shared document or a meeting tool. We assume that because the person appears to be from a known project, they are legitimate. We assume that because the app looks like a normal installer, it is safe. We have inverted the principle of 'trust but verify' into 'trust and then hope it works.'
In my years as a DAO Governance Architect—and from my own work auditing recruitment practices for Web3 organizations—I have seen a disturbing pattern. Many projects have no formal identity verification for their human resources personnel. A LinkedIn profile with a history of crypto-related posts is considered sufficient. The attack exploited exactly this gap. The attacker created a convincing recruiter persona, complete with a history of engaging with the target's posts, and then used an AI meeting tool as the pretext for the malicious installation.
What makes this attack particularly insidious is that it leverages the very narrative we champion: the promise of AI to streamline and democratize hiring. The attacker named the tool 'Relay'—a name that evokes speed, efficiency, and modernity. They leaned into the hype to lower the victim's guard. This is the dark side of narrative-driven development. When a technology becomes a buzzword, it also becomes a vector.
But the contrarian perspective goes deeper. The emphasis on 'code is law' has created a blind spot for the social layer. We are so focused on making protocols trustless that we forget to make the interactions around them trustable. The solution is not just better endpoint security—it is better community governance of trust relationships. In the same way that a DAO requires a multisig for treasury transactions, a hiring process should require a verifiable attestation from multiple existing contributors. We need to 'govern the entrance' as much as we govern the exit.
Let me be clear: I am not advocating for censorship or turning Web3 into a walled garden. I am advocating for a new type of trust protocol—one that combines cryptographic proofs with social consensus. Imagine a future where every job offer includes a signed message from the project's official wallet, timestamped on-chain. Where the recruiter's identity is tied to their ENS name, and the interview app is verified via a public registry of audited tools. Where the burden of proof shifts from the candidate to the employer.
Takeaway: Building a Resilient Community
The SlowMist report is a wake-up call. It tells us that the next frontier of Web3 security is not just about securing smart contracts, but securing human interaction. The attack on Alex is not an isolated incident; it is a signal of what is to come. As the industry grows and more capital flows into the ecosystem, the incentives for social engineering attacks will only increase. We cannot rely solely on tools like anti-virus software or hardware wallets. We must build a culture of verification, where trust is earned through cryptographic and social proof.
What does that look like in practice? First, every Web3 professional should adopt a personal security protocol for job offers: never install unsolicited software; verify the recruiter's identity through multiple independent channels (e.g., a public wallet address or a known Discord handle); use a dedicated, disposable virtual machine for any interview that requires running third-party code. Second, projects should implement on-chain recruitment verification: a simple signature from the project's treasury or multisig wallet confirming that the recruiter is legitimate. Third, the community should create and maintain a public registry of verified interview tools, analogous to how we have lists of audited smart contracts.
But beyond the practical steps, there is a deeper lesson. The attack is a reminder that the strength of Web3 lies not just in its protocols, but in its people. We are a community that values transparency, autonomy, and collaboration. We must apply these values to the recruitment process itself. We must design systems that make it easy to be honest and expensive to be dishonest. We must embed trust into the flow of human interaction, not just the flow of digital assets.
'Don't govern the exit, govern the entrance.' This principle, which I have often applied to tokenomics and DAO design, is equally relevant to human sourcing. By controlling how trust is established at the point of entry—whether it is a user joining a dApp or a developer accepting a job offer—we can prevent exploitation downstream.
The attack on Alex could have been avoided. Not by better malware detection, but by a better ritual of trust. As we move forward, let us remember that the blockchain may be trustless, but the community is not. And it is in the hands of that community to protect itself.
I recently spoke with a group of Web3 developers at a meetup in Paris. They asked me what kept me up at night. I told them: it is not hacks of smart contracts. It is hacks of human hearts. This attack is one of those. But it also shows our resilience. SlowMist's analysis, the community's rapid alerting, and the shared determination to learn from the incident are proof that we are capable of building a safer ecosystem. Not through code alone, but through conscious, caring, and connected action.
Let this article be a call to arms—not for more firewalls, but for more farsightedness. Let's design our social protocols with the same rigor we apply to our financial ones. Because code is law, but people are the soul.