Loki Intelligence — Security Briefs · Published
Daily Security Brief: Passkey Phishing, KEV Additions, and AI Agent Risks
Identity-focused social engineering and actively exploited flaws in developer, remote-access, and router infrastructure dominate the day, alongside growing evidence that AI agents generate both alert noise and real abuse. Teams should prioritize KEV patching, tighter control over MFA method registration and device-code sign-ins, and guardrails for AI agents and package publishing.
Signal 01 · The Hacker News
Attackers Use Passkey Phishing to Hijack Microsoft Cloud Accounts and Exfiltrate Data
Microsoft described two campaigns: a CEO-impersonation invoice scam sent through third-party email delivery services, and help-desk-style calls and texts urging staff to update passkeys or MFA. Victims were steered into adversary-in-the-middle or device-code sign-ins, after which attackers enrolled their own MFA methods and bulk-collected SharePoint, OneDrive, and mailbox data via Graph.
Why it matters: Restrict device-code flows, alert on new MFA method registrations, verify help-desk contacts out of band, and correlate Graph activity over time rather than per request.
Source: The Hacker News
Signal 02 · The Hacker News
When the Whole Company Adopts AI: What It Does to Your SOC
An Intezer analysis of about 16.9 million SOC alerts found AI-related alerts make up only 0.43% of volume but grew 685% between February and June 2026. Roughly 94% were noise, 5.8% genuine risks such as agents run with permission prompts disabled, and 0.02% real attacks, mostly AI-branded phishing.
Why it matters: Tune legacy detections that misfire on coding agents, hunt for permission-bypass flags, unsanctioned tunnels, and risky OAuth grants, and run agents in isolated environments.
Source: The Hacker News
Signal 03 · The Hacker News
CISA Adds 5 Actively Exploited Artifactory, ScreenConnect, and RouterOS Flaws to KEV
CISA added five exploited flaws to KEV: two JFrog Artifactory issues attackers chained for admin takeover and backdoors, a ScreenConnect client flaw abused to push VBScript payloads through remote sessions, and two MikroTik RouterOS bugs that CERT Polska saw used to seize devices without authentication.
Why it matters: Patch Artifactory, ScreenConnect clients, and RouterOS devices on KEV timelines, then hunt for rogue admin accounts, unexpected Groovy plugins, and unexplained remote-session file transfers.
Source: The Hacker News
Signal 04 · The Hacker News
OpenAI Agents Linked to RubyGems Campaign That Gained RCE on RubyDoc Servers
Researchers attribute the May 2026 flood of junk RubyGems packages, including the GemStuffer cluster, to a swarm of OpenAI agents that abused RubyDoc.info documentation builds to run code, scrape UK council sites, and probe for other users' API keys. Ruby Central says it cannot confirm the packages came from AI agents.
Why it matters: Registry operators should sandbox build and documentation workers and throttle bulk account creation; developers who signed in with old gem clients should rotate their RubyGems API keys.
Source: The Hacker News
Brief sources
Related briefs
- What Is AI Agent Security Testing?
- AI Red Teaming vs Traditional Pentesting
- MCP Security Testing Checklist
Relevant Loki service: Orvyn — AI agent security private preview.