The first sign was a LinkedIn message from a recruiter at a well-known Web3 venture fund. The profile photo was professionally bland, the headline filled with buzzwords about ‘DeFi scaling solutions.’ They invited me to an AI-powered interview using a tool called Relay—a ‘next-gen video platform for remote hiring.’ I clicked the link out of curiosity. The installer asked for permissions that no interview software should ever need: access to my browser keychain, read my Telegram session files, scrape my crypto wallet directories. I stopped. But thousands of others, hungry for the next opportunity in a bull market, did not. This is not just a scam; it is a systemic failure of trust at the very interface where Web3 meets the human layer.
Context: The Anatomy of the Relay Attack On July 29, 2025, SlowMist released an analysis of a new info-stealing malware campaign targeting Web3 professionals. The attack vector is pure social engineering: attackers impersonate recruiters on professional networks like LinkedIn, send a link to a fake AI interview tool called Relay, and trick victims into installing a cross-platform malware that exfiltrates browser credentials, cryptocurrency wallet data, macOS Keychain contents, and Telegram session tokens. The malware exists for both macOS and Windows, indicating a sophisticated team capable of reverse-engineering operating system security models. The payload appears as a legitimate installer but silently drops a stealer that communicates with command-and-control servers to siphon data. SlowMist obtained a sample and confirmed the full kill chain.
What makes this attack uniquely dangerous is its timing and target. We are in mid-2025, a bull market where every Web3 native is chasing the next role at a high-flying protocol or VC. AI tools are ubiquitous—everyone expects a video interview with a chatbot or an automated assessment. The attackers are leveraging that normalization. They are not breaking cryptographic primitives; they are exploiting the gap between our trust in technology and our vigilance against human deception. This is the same gap that has enabled every phishing scam since the 1990s, but now with a Web3 twist: the prize is not just a bank password but a seed phrase that controls a lifetime of savings.
Core: The Paradox of Transparency in a Cashless Society From my perspective as a CBDC researcher in Lagos, I have watched the promise of financial inclusion collide with the reality of predation. In 2017, during the ICO boom, I recorded how hyperinflation drove Nigerians to Bitcoin not out of greed but out of survival—they needed an asset that could not be debased by a corrupt central bank. Yet even then, the same survival instinct made them vulnerable to scams. Today, the pattern repeats with a new costume: the AI interview. The attackers understand that Web3 professionals are conditioned to trust decentralized tools and open-source software. They mimic that ethos by branding Relay as a “community-built interview platform” on GitHub (a fake repository, of course).
The data tells a story of targeting precision. According to the SlowMist report, the malware specifically queries for directories related to MetaMask, Phantom, Keplr, and over 40 other wallet extensions. It also scrapes iCloud Keychain on macOS, which often stores passwords for centralized exchanges and email accounts. The Telegram session theft is especially insidious—attackers can use the stolen session to impersonate the victim inside private Web3 groups, launching secondary phishing attacks against colleagues who might trust a known username. This is not random spraying; it is a surgical strike on the most valuable information in the Web3 ecosystem.
During my work analyzing the Nigerian CBDC (eNaira) pilot, I reverse-engineered its offline transaction layer and discovered a vulnerability in the card issuance logic. The flaw was not in the cryptography but in the user enrollment process—people were trusting agents who had no incentive to verify identities properly. Similarly, the Relay attack exploits trust in the hiring process itself. The solution is not just better antivirus software; it is a fundamental redesign of how we verify identity and intention in a decentralized world. The silence between transactions—the moments when a user decides to trust or not trust—is where the real security battle is won or lost.
Contrarian: The Decoupling Thesis—Why Code Is Not Enough The prevailing narrative in Web3 is that code is law, and that smart contracts eliminate the need for trust. But this attack proves the opposite: no matter how secure the blockchain layer is, the human interface remains the weakest link. The contrarian angle here is that the very tools we are building to decentralize hiring—like DAO-based recruitment platforms or on-chain reputation systems—may actually increase the blast radius of such attacks. If your entire identity is aggregated into a set of on-chain credentials (like a Decentralized Identifier wallet), a single compromise can revoke your access to every application, every job, every community you belong to.
Moreover, the market’s reaction to such security events reveals a blind spot. In a bull run, FOMO drowns out caution. Projects that promise to solve this problem—like ZK-based identity verification or secure enclave browsers—will attract hype and funding, but their adoption will be slower than the evolution of the attacks. We are in a race where the attackers are iterating faster than the defenders because they are human-centered, while defenders are code-centered.
Let me offer a personal data point: during my 2020 analysis of DeFi Summer’s predatory lending practices, I observed that the protocols with the highest TVL were often those with the most generous yield farming rewards—but also the most opaque social engineering of users. The same psychological principle applies here: the promise of a high-paying Web3 job lowers resistance to installing unverified software. The decoupling thesis—that crypto assets can decouple from traditional financial risks—fails when the risk is not financial but psychological.
Takeaway: Rebuilding Trust in the Empty Spaces What does this mean for the next market cycle? The Relay attack is a harbinger of a broader trend: as AI tools become indistinguishable from human interaction, social engineering attacks will become automated and personalized at scale. We will see deepfake recruiters, AI-generated code review sessions that install backdoors, and simulated hackathons that steal credentials. The industry must shift its investment from pure code audits to live, real-time human behavior verification.
I propose a structural alternative: zero-knowledge interview environments where candidates run only in isolated virtual machines that sync to an immutable ledger of actions, but never expose the host system to malicious payloads. Additionally, we need privacy-preserving reputation proofs that allow a recruiter to verify a candidate’s credentials without the candidate ever revealing their private keys or personal data. The technology exists—we only lack the will to prioritize it over short-term hiring speed.
The paradox of transparency in a cashless society is that we must give up some privacy to gain security—but only if that transparency is designed with empathy for the human user. Listening to the silence between transactions means recognizing that the most important transaction is the one where a person decides to trust. That decision cannot be automated away. It must be secured by systems that are as compassionate as they are robust.
In my Lagos-based research, I learned that the unbanked do not need just technology; they need guardians who understand the liminal space between code and care. The Relay attack reminds us that Web3 is not a trustless utopia—it is a trust migration from institutions to individuals. And individuals, without the right tools, are fragile. The question we must answer before the next wave of attacks is not "how do we build better code?" but "how do we build better trust?"