Abnormal AI

Abnormal AI is joining CrowdStrike's Project QuiltWorks, which surfaces vulnerabilities at the scale and speed of frontier AI. Frontier AI has changed the speed and scale at which attackers can move. Project QuiltWorks addresses that shift by uniting AI-driven discovery, adversary-informed prioritization, remediation through a global network of systems integrators, financial protection through leading cyber insurers, and cloud infrastructure hardening with AWS into a single framework for frontier AI risk. As a longstanding CrowdStrike partner, Abnormal extends and deepens that framework by adding behavioral security, closing exposure gaps other tools miss and giving mutual customers a more complete answer to where they're exposed. It builds on our existing integrations with Falcon Identity Protection, Next-Gen SIEM, and the Malware Analysis Agent. Read more: https://lnkd.in/ez6YV2ZX

This week: over-provisioned new hires, a dormant account waking up, employees selling their own access, AI features vendors switch on without asking, and a CFO's voice that wasn't real. Attackers don't need to break in when the access is already sitting there unwatched, or when a familiar voice alone still passes as proof. The only place any of these show up is behavior, which is exactly what static access and identity controls were never built to see. Featuring contributions from: Nagendra Kumar Gurram Stephen Harrison Abhishek Anbazhagan Tanisha Nalavadi Hyperdrive is Abnormal's weekly newsletter sharing how our product and security teams think about building AI-native security — a compilation of daily posts on the attacks reshaping our roadmap and the gaps that haven't been closed yet. Subscribe to see how and what we're building.

AI Phishing Coach now trains employees differently based on how they actually respond to threats, not a fixed calendar. Completion rates are high, click rates are low, but risky behaviors keep showing up in incidents. The problem is that most security awareness programs were built to be measured, not to change behavior. This update changes that: 🎯 Precision Training automatically adjusts simulation frequency and difficulty based on each employee's actual behavior patterns. 📱 QR Code Simulations close the gap people don't know they have. 🔁 In-Thread Follow-Up re-engages employees who ignore a simulation. ✍️ Attack, Training & Campaign Creator lets admins generate custom simulations and training videos from a plain-language prompt. Details are in the link below. Find us at Fal.Con next week to demo AI Phishing Coach. https://lnkd.in/ePgdhdWK 

OpenAI is calling for collective action on cyber defense, and Abnormal is proud to stand behind it. AI is changing the entire threat landscape. Much of the conversation about AI-enabled cyber risk has focused on software vulnerabilities and malware. That's only part of the picture. AI also gives attackers new ways to exploit people, identities, trusted relationships, and business processes, from more convincing social engineering to identity compromise and supply chain fraud. Stopping AI-driven threats requires behavioral AI security: systems that learn what normal looks like for every identity in an organization and catch what deviates from it, going far beyond traditional approaches focused on known-bad signatures. Bad AI will keep getting better. Good AI has to get better faster. Learn more in our blog from our CEO and Founder Evan Reiser: https://lnkd.in/ePuFa_-P

Control Center gives security teams precise rules and custom AI models, layered on top of Abnormal's behavioral AI. Not every security decision is behavioral. A vendor's domain should be safelisted unless its mailbox is compromised, and a harassment campaign should be flagged even when the message looks clean. Until now, that meant hand-maintained detection logic running alongside your AI. Control Center removes that tradeoff. Teams can define precise conditions with Custom Rules or describe a pattern in plain language with Custom AI Models. No detection engineering required. Full details at the link in comments. Find us at Fal.Con next week for a demo!

Cyber attacks are becoming harder to distinguish from legitimate business activity. In this episode of Today in Tech, Abnormal AI Security's Piotr Wojtyła joins host Keith Shaw to explain how attackers use AI, vendor impersonation, compromised accounts, and business context to make fraudulent requests appear routine. If employees can no longer rely on poor grammar or suspicious links as warning signs, what comes next? Watch the episode: http://spr.ly/6041B1xpFV #TodayinTech #Cybersecurity

We're enhancing the Abnormal Behavior Security Platform across three critical surfaces of email risk: what's coming in, what's going out, and the people attackers target. Today's most effective attacks look like normal business activity, which is exactly why defending against them now takes more than inbound detection alone. 💻 Control Center: Turn organization-specific knowledge into detection decisions through precise rules and custom AI models, with full visibility into whether each decision came from core AI, a custom model, or a custom rule. 📩 Email DLP Rules: Combine explicit outbound policies with an AI Triage Agent that reviews every match in context, auto-clears false positives, and logs every verdict. 🎣 AI Phishing Coach: Now calibrating simulation frequency and difficulty based on how each employee is actually targeted, not a fixed calendar. Full details on all three in our latest blog https://lnkd.in/e3fhU9cs and find us at Fal.Con next week for a demo!

Using AI to work faster is not the same as transforming how work gets done. Lisa Wallace, our Chief People & Transformation Officer, breaks down the difference: 💡 A technology switch is plug-and-play. You use an LLM to write better feature requests and send them to your product team faster. Useful, but incremental. An AI transformation is a step-function change. Every call recorded automatically. Customer insights mined without anyone manually pulling them. Features prioritized and routed without a human in the loop at every step. Same goal. Completely different level of impact. See how we're transforming at the link in comments.

McLaren Construction Group adopted Abnormal to solve one problem: too much junk email flooding executive inboxes. Today, they've saved 196 hours on graymail management and 41 hours on user-reported messages per month. Hear directly from Dan Blackman, Group IT and Security Director at McLaren Construction, on how Abnormal eliminated the noise and became the foundation for stopping modern threats. Read the full story at the link in comments.

This week: AI copilots as insider accomplices, the pre-resignation window nobody's watching, guest identities nobody offboards, and the access that outlives every role change. Every one of these gaps has the same shape: access that was granted once and never questioned again. The behavior gives it away long before the entitlement review does. Featuring contributions from: Tanisha Nalavadi Vivek G. Shaun Archer Akanksha Saxena Hyperdrive is Abnormal's weekly newsletter sharing how our product and security teams think about building AI-native security — a compilation of daily posts on the attacks reshaping our roadmap and the gaps that haven't been closed yet. Subscribe to see how and what we're building.