SlowMist just dropped the analysis. The attacker didn't break any code—they broke your trust.

A new piece of malware disguised as an AI meeting tool called 'Relay.' Target? Web3 professionals actively job-hunting. The payload? Full access to your crypto wallets, browser credentials, Telegram sessions, even macOS Keychain. And it's cross-platform—Windows and macOS.

This isn't your generic phishing email. This is a precision strike against the one vulnerability no smart contract can fix: the human who clicks 'Install.'

Context: Why Now?

We're in a bull market. Hiring in crypto is exploding. AI tools like Otter.ai, Fireflies.ai, and custom meeting bots are standard. Every remote worker expects to download a 'meeting scheduler' or 'transcription plugin.' Attackers know this.

The 'Relay' malware piggybacks on the AI narrative. The recruiter (fake LinkedIn profile) sends a Calendly link. You book. They send a pre-meeting software requirement: 'Please install our AI assistant, Relay, to sync your availability.'

It looks legitimate. Maybe you've done this before. But Relay is a custom info-stealer, built from scratch. No open-source repo. No public domain. Just a binary that, once run, starts extracting everything.

This is not theoretical. SlowMist's threat intel team already extracted and analyzed the payload. The IOCs are public. But the damage is already done for the first wave of victims.

Core: The Technical Forensics (What I Actually Looked At)

From my audit experience, the first thing that struck me was the target scope. This malware doesn't just grab browser cookies. It specifically targets:

• Browser credential stores: Saved passwords, autofill data, session tokens.
• Cryptocurrency wallet extensions: MetaMask, Phantom, Keplr, Rabby—any Chrome-based wallet with a 'prefs' file. It reads the encrypted seed phrase storage path.
• macOS Keychain: All entries—including API keys, SSH passwords, and any saved wallet mnemonics.
• Telegram sessions: The tdata folder. Full access to account—read messages, impersonate, spread the same malware to your contacts.

That last one is critical. Even if you use a hardware wallet for cold storage, your Telegram session gives the attacker your social graph. They can DM your CTO, your VC, your co-founder. The infection chain continues beyond you.

The code is simple but effective. No advanced persistence (no rootkit, no DLL injection). It relies on the victim being 'tricked once.' But that's the genius: it doesn't need to be stealthy. It just needs to run once.

Based on my monitoring, the malware calls home to a C2 server that resolves to a bulletproof hosting provider in Eastern Europe. The stolen data is exfiltrated via HTTPS POST—no encryption beyond TLS. The attackers likely parse the credentials manually or sell them in bulk.

And yes, I verified: the binary is signed with a stolen or self-signed certificate dated July 2025. It passes the macOS Gatekeeper check for now—Apple hasn't revoked it yet.

Contrarian: The Blind Spot Everyone Misses

The mainstream take is: 'Be careful with downloads.' That's useless advice. The real insight? This attack reveals that trust in 'AI tools' has become a new attack surface—and security vendors are not tracking it.

Think about it. Every Web3 security conference talks about smart contract audits, MEV protection, and hardware wallets. Nobody audits the HR tech stack. But the attacker just found the weakest link: the pre-hire software pile.

Here's the contrarian angle: Even if you never install the malware, the social engineering alone is dangerous. The fake recruiter already has your resume, your LinkedIn profile, your email. They know you're looking for a job—which means they know your current employer might not have great security. The spear-phishing doesn't need the payload to succeed. A well-crafted fake offer letter with a malicious PDF can achieve the same.

The second blind spot: Multi-sig setups are not safe from this. If the attacker steals the Telegram session of one signer, they can monitor internal chats. They see when a transaction is being prepped. They can time a compromise when the human guard is down.

I've seen this before in 2023 with the 'Lazarus Group' style attacks targeting DeFi developers. But that was state-level. This one feels different—smaller, faster, more accessible. Any script kiddie with $5,000 can buy this malwarized AI tool on the dark web.

Takeaway: What You Should Actually Do

The window of inaction is closing. Apple hasn't revoked the certificate yet. LinkedIn hasn't purged the fake profiles. The C2 server is still live.

Here's my forward-looking judgment: Within the next 30 days, we will see at least three copycat variants using different AI brand names (Otter, Fireflies, etc.). If you or your team is hiring, do this now:

1. No unsolicited software installations. Period. If a recruiter asks you to install a tool before the first call, demand a phone call verification instead.
2. Use a dedicated device for job interviews. A cheap Chromebook or a VM that you wipe after 48 hours. No personal wallets, no Telegram sessions.
3. Check your Telegram session list. If you see an unknown device, kill it immediately and rotate your 2FA.
4. Monitor the SlowMist IOC feed. Block the known C2 domains on your corporate firewall.

This isn't a warning—it's a forensics report. The evidence is on the table. How you respond decides whether your next interview is a career move or a total loss.

Disclaimer: This analysis is based on publicly available threat intelligence from SlowMist and my own technical review. I do not own or control any of the mentioned assets. Always verify software from recruiters through official channels.