ai
The Next AI Security Frontier: “Agents With Hands” Are Becoming a Board-Level Risk
Your new “AI helper” is basically shadow IT with hands 🤖🧨 Untrusted content → model decides → tools execute. That’s the breach loop.
ai
ai
If your “AI Coworker” Gets Targeted, What Tips You Off First?
Your “AI coworker” isn’t the breach. The OAuth trust event is. 🔥🕵️♂️ Device-code phishing + consent traps = “approve to exfil.” (And yes, AI agents are already being used as the wrapper.)
PIR
Deepfake BEC & Payment Diversion: The Q1 2026 Fraud PIR You Can’t Defer
Deepfake BEC = the same old fraud… with a way better script. 🎭💸 If payroll/AP changes can happen on “sounds right,” you’re funding someone’s Q1 bonus.
![[FORECAST] CoPhish: The Microsoft Copilot Link That Hands Over Your OAuth Tokens](/content/images/size/w600/2026/01/z-1.png)
forecasts
[FORECAST] CoPhish: The Microsoft Copilot Link That Hands Over Your OAuth Tokens
Will at least one publicly disclosed enterprise breach be confirmed where attackers used a Microsoft Copilot Studio..
edr
CrowdStrike vs Microsoft Defender: Who Leads EDR/XDR Into 2026?
EDR “leader” in 2026 = who contains fastest at scale + doesn’t implode during updates. 🎄🧯 Our model: CrowdStrike 50% (±8), Microsoft Defender 35% (±7), SentinelOne 15% (±5).
ai
Dark LLMs: When Your AI Traffic Is C2
Your “normal” AI traffic can be stealth C2 now. Dark LLMs are writing per-host pwsh one-liners, self-rewriting droppers, and hiding in model APIs you approved. If you’re not policing AI egress, you’re not doing detection. 😬🤖
forecasts
AI Agents as Regulated C2: Will Anyone Be Forced to Act?
AI just ran most of an espionage op, and regulators are still in “interesting case study” mode. 😏 We’re forecasting: 55% odds that by 2026, someone will force signed AI connectors + agent logs by default.
ai
Your AI Agents Are the New C2 — Lock Down Identity & Connectors
Anthropic just showed what happens when your “helpful” AI agents become C2: 80–90% of an espionage op automated, humans just clicking approve. Lock down identity + connectors or you’re renting your SaaS to someone else’s botnet. 🤖🚨
weekly
SIGNALS WEEKLY: The Quiet Shift- When Intrusions Start Thinking for Themselves
A Chinese crew let a jailbroken AI run most of the intrusion while FortiWeb + Firebox burn in KEV and a contractor leak drops the playbook.
unc3886
Vishing Meets Cloud: UNC6040’s Abuse of Salesforce Connected Apps for Stealthy Data Exfiltration
UNC6040 is a financially motivated threat actor specializing in voice phishing (vishing) campaigns that abuse Salesforce Data Loader connected apps to gain unauthorized access and exfiltrate sensitive data. This novel attack vector leverages social engineering via telephone impersonation of...
phishing
AI-Driven Voice Deepfake Defense: Integrating Detection and Watermarking into Healthcare SIEM and SOC Workflows
Healthcare organizations with SIEM deployments and immature SOCs face escalating risks from AI-driven vishing attacks leveraging voice deepfakes. This analysis outlines a pragmatic, phased approach for integrating AI-based voice deepfake detection and audio watermarking..
m-trends
AI-Driven Cyber Threats, Ransomware Evolution, and Supply Chain Security: We (try to) PREDICT what's coming in Mandiant's 2025 M-Trends Report
I take a SPECULATIVE deep dive into what I think might be in the 2025 Mandiant M-TRENDS report.