Book a demo →
Series A
$36M led by Battery Ventures, with Accel and Foundation Capital
Blog

Top Threat Research blog posts

Research, attack teardowns and product notes from the AegisAI team.

Topics

All postsEmail SecurityAIThreat ResearchTechnical GuidesAnnouncementsCustomer StoriesPress Releases
September 23, 2026

Red Tide: A Calendar-Invite Worm That Spreads Mailbox to Mailbox

A phishing campaign that spreads itself: in one confirmed case, a personal Gmail account picked up the lure, and nineteen minutes later a different mailbox at the same company was sending it onward.

Read article →
September 18, 2026

The Inbox Has a Second Reader: Prompt Injection in Gemini and Copilot

Attackers hide instructions in email that Gemini and Copilot read and that people cannot see. How email prompt injection works, and how Aegis catches it.

Read article →
Diagram showing an AI agent registering domains and rebuilding malware on a loop, a victim entering a device code at a genuine Microsoft sign-in, and the resulting access and refresh token landing with the attacker
September 11, 2026

Midnight Blizzard-Linked Actor GTG-20006 Automated Device Code Phishing With AI

Anthropic says GTG-20006, a Midnight Blizzard-linked actor, used AI to automate device code phishing against 20+ government and defense organizations.

Read article →

All Threat Research posts

Adversary-in-the-middle diagram. A victim logs in through a reverse proxy to the real login page and passes MFA. The proxy copies the session token to the attacker, who is logged in without a password.
September 8, 2026

Email Security Threats in 2026: BEC, AiTM, and Zero-Day Phishing Explained

The 2026 email threat landscape: BEC, vendor email compromise, AiTM session theft, and zero-day phishing, the attacker tooling behind them, and why legacy filters miss them.

Read article →
Google Calendar phishing loop: a compromised mailbox sends a DKIM-signed invite that auto-creates an event, clears a single-use gate, and installs an RMM agent on the attacker's tenant, making the victim the next sender. AegisAI threat intelligence.
September 2, 2026

Google Calendar Phishing: The Invite That Installs an RMM Agent

Malicious Google Calendar invites rose tenfold in a week. Inside a campaign that auto-creates events from stolen mailboxes and installs a signed RMM agent.

Read article →
AI-generated phishing passing through two gates: a tier-1 Gmail and Microsoft filter that delivers 50.3 percent of it against 28.5 percent for human-written, then the recipient, where over 60 percent click, ending in compromised credentials.
September 2, 2026

Half of AI Phishing Reaches the Inbox. Then 60% of People Click.

CrowdStrike puts AI phishing click-through past 60%. Our analysis of 20,000+ emails shows why it arrives: 50.3% clears Gmail and Microsoft filters.

Read article →
Diagram of the Mirage2FA relay: a victim's Microsoft 365 login passes through the attacker proxy, MFA is satisfied, and the session cookie is siphoned into an attacker session.
August 28, 2026

How Mirage2FA bypasses Microsoft 365 MFA with stolen session cookies

Yes, attackers bypass Microsoft MFA. The Mirage2FA AiTM kit relays the real login and takes the session cookie the moment MFA clears. Here is how it works.

Read article →
IEH credential phishing attack path: an impersonated email passes reputation, attachment, and signature filters with no match, reaches a counterfeit Microsoft login page, and ends in a compromised Microsoft 365 mailbox.
August 17, 2026

Credential Phishing With No Payload: Inside the IEH Microsoft 365 Mailbox Breach

A credential phishing email with no malware opened a Microsoft 365 mailbox at IEH Corporation, a U.S. defense supplier. How the account takeover worked.

Read article →
July 31, 2026

Kali365: How a Device-Code Phishing Kit Defeated MFA Without Breaking It

A $250-a-month phishing kit turned a legitimate Microsoft sign-in feature into a way to steal Microsoft 365 access without a password, a credential form, or a single attacker-owned domain, then staged its own shutdown and kept running under a new name.

Read article →
TIDALGUEST self-replicating phishing worm: a stolen inbox becomes the next trusted sender, spreading through contact lists. AegisAI threat intelligence.
July 4, 2026

TIDALGUEST: A Self-Replicating Invitation Phishing Cluster

TIDALGUEST is a self-replicating invitation phishing cluster that turns each stolen inbox into a new sender and layers five evasion techniques past URL scanning.

Read article →
April 22, 2026

AegisAI Threat Intelligence Series: How to Bypass a Secure Email Gateway Part 1

Backscatter phishing is an emerging threat where attackers exploit legitimate Non-Delivery Reports (NDRs) to bypass Secure Email Gateways (SEGs). By sending emails to non-existent addresses with a spoofed From field, attackers force legitimate mail servers to bounce a notification, containing the original malicious payload, directly to the intended victim. Since these bounce messages originate from trusted, RFC-compliant Mail Transfer Agents (MTAs) and pass SPF and DKIM checks, they are often invisible to traditional security tools.

Read article →
The New World of AI Spearphishing — 2025 FBI IC3 report analysis
April 20, 2026

The New World of AI Spearphishing

Attackers stopped writing malware. They started writing emails. This paper breaks down the $11.64B shift from payload-based attacks to identity exploitation — and why your existing stack can't see it coming.

Read article →
February 15, 2026

AI Security Threats Ranked By What's Actually Exploited

Part 2 of 6: AI Security Market Analysis Series

Read article →
February 11, 2026

Aegis Threat Intelligence: Sharpening the Harpoon – Direct Send Abuse Still Fueling Modern Spearphishing

The call is coming from inside the house: How threat actors act as 'internal' users to bypass your security stack.

Read article →
January 25, 2026

Operation Social Undertow: A Phishing Campaign Spoofing the Social Security Administration

Threat actors deploy SimpleHelp RAT via sophisticated SSA phishing.

Read article →
December 14, 2025

The Bullseye Report: Criminal Account-Based Marketing (ABM) Attacks and the New VIP Risk

Why C-Suites Face 51% of All AI Attacks

Read article →
December 9, 2025

The Anatomy of an AI Attack

Anatomy of a Hack: How AI Clones Your Company in 3 Steps

Read article →
December 3, 2025

AI Email Security: Why ROI Shows Up Here First

AI agents are finally delivering real security outcomes. The first place that shows up? Your inbox.

Read article →
December 2, 2025

The AI Supply Chain You Can’t See: Mixpanel, OpenAI, and the Risk of Third-Party Model Exposure

A smishing attack at Mixpanel exposed OpenAI API user metadata and blew open a bigger question: which of your vendors are quietly sending data to OpenAI or custom models? Break down the hidden AI supply chain and what to do about it.

Read article →
December 1, 2025

S-Curve of Crime: Why AI Phishing Growing 100% YoY

Phishing click rates are up 190% in 2024. New research from Aegis AI reveals why AI-powered attacks are following the "S-Curve" of SaaS adoption.

Read article →
November 26, 2025

The 2025 Holiday Heist: 5 AI-Powered Scams Targeting Employees This Black Friday

From deepfake ads to 'HR bonus' phishing, AI has changed the game this holiday season. Here are the 5 threats IT Admins need to warn their teams about right now—and the corporate risks attached to each.

Read article →
September 14, 2025

Trusted but Not Verified: A Case Study of Compromised Infrastructure Spearphishing Attacks

New research analyzes the "UN Summit" phishing campaign, a textbook example of "Living off the Land" attacks that evade rule and reputation based detection. With 40% of identified government impersonation attacks now originating from spoofed or compromised domains, this post explains why Artificial Intelligence analyzing emails like an analyst is the only viable defense against the next generation of AI Spearphishing.

Read article →

See what AI-native security feels like

Thirty minutes. Real attacks pulled from environments like yours (BEC, vendor fraud, credential phishing) with the reasoning behind each verdict. No setup required.

Request a demo
→

Five-minute connect · monitoring mode · disconnect any time without touching mail routing