184 million login credentials exposed

API keys, unencrypted GCP storage, and nearly 900K users exposed—what went wrong and how to prevent it.

June 12, 2025

Hey Security-Minded Reader 👋

You’re reading the very first issue of this newsletter—a new dispatch at the intersection of cybersecurity and artificial intelligence.

If you care about breaches, AI-driven threats, security protocols, and how all of it is evolving faster than regulators or legacy systems can keep up—you're in the right place.

🛡️ IN TODAY’S EDITION

🧨 1. Breach of the Week

  • Target: Google & Apple (plus Microsoft, Facebook, Instagram, Snapchat, and others)

  • Vector: Unprotected database exposed 184 million login credentials

  • Impact: 184,162,718 unique usernames and passwords leaked

Lessons:

  • Infostealer malware remains a massive threat, siphoning credentials and sensitive browser data from infected devices

  • Databases containing sensitive information must be encrypted and access-controlled—no exceptions

  • Users should regularly change passwords, enable multi-factor authentication, and avoid storing sensitive documents in email inboxes

🧠 2. AI Accelerates Government App Breach

Hackers breached TeleMessage—a Signal-based app used by US officials—by exploiting AWS misconfigurations to steal admin credentials and government communications data.

💡 AI-powered reconnaissance tools are shrinking the attack window, automatically scanning for exposed cloud assets faster than defenders can patch them.

📉 Even encrypted messaging apps fall victim when backend infrastructure lacks proper security.

📍Takeaway: AI-driven attacks require AI-powered defenses and automated monitoring.

🔒 3. Protocol News: CISA AI Security Guidance

CISA and international partners released fresh guidance this May for securing AI systems from training to deployment.

💡 Key focus: data supply chain security—protecting the data that feeds AI models from tampering and unauthorized access.

📉 Addresses critical risks like malicious data modification and "data drift" that can silently corrupt AI performance over time.

📍Takeaway: Review your AI deployments now—especially access controls, logging, and data integrity measures.

📡 4. Attack Surface: IoT & Cloud

Recent breaches (Mars Hydro, TeleMessage) reveal how sprawling modern attack surfaces have become:

💡 IoT devices like hydroponic controllers leaked billions of records—Wi-Fi passwords, IP addresses—due to terrible security defaults.

📉 Cloud misconfigurations get exploited within minutes by automated bots scanning 24/7 for vulnerabilities.

🔓 Even "secure" messaging platforms crumble when backend controls are weak.

📍Takeaway: Every device, API, and cloud bucket is now an endpoint. Harden them with perimeter-level rigor.

🔓 5. Free Resources for You

Here’s what I’ve found most helpful this week:

👉 For Now, One Quick Question

Since this project is just getting started, I’d love to hear from you early!

What’s keeping you up at night in Cyber or AI?

Login or Subscribe to participate in polls.

💬 Hit reply and let me know—I'll build this newsletter to serve the challenges you're facing, not just the ones trending on Twitter!

🔐 Stay sharp. Stay secure.
This newsletter is crafted with focus, skepticism, and zero hype—just field-relevant insights at the intersection of cybersecurity and AI.

💬 Got a tip, tool, or topic suggestion? Hit reply—I read every message.
📎🌍 Published by Sentinel | [LinkedIn]