SentinelOne

AI agents now hold credentials and act across the enterprise. At OneCon, we unveiled: 🛡️ Singularity™ ADR to stop rogue and malicious agents at machine speed ⚡ Singularity ASOC brings security superintelligence to the SOC 🤝 An expanded PartnerOne program to equip partners to build and scale AI Security practices Learn more 👉 s1.ai/mtbf7

Gemini 4 Carbon is now available, and our security research team evaluated Carbon on kernel driver vulnerability research against other frontier models. It uncovered four verified vulnerabilities beyond the core memory-corruption bugs every model found. Learn more 👇

AI agents are taking action across the enterprise. From visibility to runtime guardrails, SentinelOne SingularityTM Agentic Detection and Response (ADR) shows what agents are doing and enforces the limits you set. Watch the demo. 👇 youtu.be/NyrVzRL3_aQ?si=eQWQ…

🚨 This week, law enforcement indicted two Canadians over a university hacking scheme and seized over 100 investment fraud across Europe. Attackers posed as IT support over Microsoft Teams and refrigeration systems are being disrupted in a second wave of attacks. bit.ly/4kvWVYo

As a longtime participant in Daybreak Defense Network, SentinelOne has played an active role in testing, benchmarking and shaping the latest Astra-class frontier models, like GPT-6 Cyber. "The hardest part of security research is connecting scattered clues into a clear understanding of how an attack could work. GPT‑6 Cyber helps our researchers make those connections, test their hypotheses, and uncover weaknesses that demand deeper investigation. It expands what our experts can take on, and helps us turn that understanding into better protection for customers.” – Chris Corde, Chief Product Officer at SentinelOne. Learn more 👇

The City of Los Angeles secured 20,000+ endpoints across 40+ departments with SentinelOne in under six months. Wayfinder MDR adds the capacity of nine full-time employees, and workflows now run in seconds instead of hours. sentinelone.com/customers/ci…

AI can reverse engineer malware. We'd still like to see the receipts. @LabsSentinel developed VerifiedIDA to keep LLM-driven analysis tied to evidence, open to review, and reusable across security workflows. Explore the full research: sentinelone.com/labs/verifie…

Threat actors are turning up the heat on Europe's supermarkets. Literally. Refrigeration systems across Europe have been disrupted, with attackers suspected of manipulating temperature controls through exposed management interfaces. Read the analysis: sentinelone.com/labs/defrost…

AI could give defenders a bigger advantage than attackers. Steve Stone made that case at OneCon. PHP2SHELL offers a recent example, with AI-assisted research helping uncover critical PHP vulnerabilities that might otherwise have gone unresolved. 🔗 bit.ly/4AcTGKb

We’re committed to defending the surfaces AI runs on. Today, we announced an integration with Anthropic’s @claudeai Cowork to extend AI Security coverage to one of the most popular agentic work apps. Runtime security will run inside the Cowork VM so security teams can: → See what agents do → Get alerts on suspicious behavior → Autonomously stop attacks Get the details: s1.ai/5351bc

AI research keeps pushing for more powerful models. Harvard professor James Mickens thinks it's time to rethink that priority. At LABScon26, Mickens made a provocative case: right now, the only good AI research is safety research. Read: bit.ly/4BDWYHF

🚨 Various vulnerabilities and security events took over this past week in what’s being called “Black October”, which attackers reportedly chaining flaws to root firewalls and cyber defenders releases emergency patches for Node2SHELL. 🔗 bit.ly/3W8UgtH

The SentinelOne plugin is live in the Plugin marketplace in Google @Antigravity! 🔌 Connect Purple AI to Antigravity agents to query Singularity Console, triage vulnerabilities next to the code, investigate alerts, and hunt threats in natural language. s1.ai/0520a4

#AI is finding vulnerabilities faster than teams can triage them. Wayfinder Frontier AI Services is adding Claude #Mythos 5.1, Anthropic’s most capable model class for cybersecurity, to find what exists, validate what’s exploitable today, and set the order for fixes. s1.ai/e83eac

With PHP2SHELL, eMudhra, LiteLLM and more, AI is helping attackers move faster and cheaper. That was the message at #OneCon: fundamentals and defense in depth win, and behavior is the signal that holds up ⚡ Join us on Wednesday for a deeper look into the threats we are monitoring: s1.ai/32cef5

🚨 This week, a Nigerian man was sentenced to eight years for spearphishing and tax fraud, while a China-linked actor deployed agentic AI to compromise cloud supply chains and PHP warns of two critical-level RCE flaws. 🔗 bit.ly/4i3g4PQ

PHP2Shell: two critical PHP flaws (CVSS 9.2) let unauthenticated attackers execute code on web servers. Patch first. SentinelOne stops post-exploitation activity at runtime in protect mode. s1.ai/09aa56