Yesterday, I have shared on LinkedIn, how you can investigate user's sign in history. I will share a link to KQL query, which will tell you - What IP addresses user used in the past 7 days - If he signed from them in the past 90 days - Did he sign from the country? Was device managed? If you are investigating user's activity, it's a great first check you can do and very useful for AI agents during incident response. https://github.com/Pavel-Hrabec/KQL-Queries/blob/main/Sign-In%20Login%20History

One thing is becoming increasingly clear: the impact of frontier AI models like ChatGPT 5.5 and Claude Mythos is now reshaping cybersecurity at every level. This cycle highlights one of Microsoft’s largest Patch Tuesdays ever, renewed concerns around AI-powered operating system features and the growing pressure on defenders to adapt to machine-speed threats. 🚨 Microsoft Patch Tuesday Fixes 167 Flaws & 2 Zero-Days What happened: Microsoft released one of its largest updates ever, fixing 167 vulnerabilities, including two zero-days — one actively exploited in the wild. The most critical is CVE-2026-32201, a SharePoint spoofing flaw that allows attackers to manipulate data and access sensitive information. 🔗 https://www.bleepingcomputer.com/news/microsoft/microsoft-april-2026-patch-tuesday-fixes-167-flaws-2-zero-days/ 📂 13.5 Million McGraw Hill Accounts Leaked in Data Breach What happened: The ShinyHunters group leaked data from 13.5 million McGraw Hill users, reportedly due to a Salesforce misconfiguration. The breach exposed personal data and highlights ongoing risks tied to SaaS misconfigurations and weak cloud access controls. 🔗 https://www.bleepingcomputer.com/news/security/data-breach-at-edtech-giant-mcgraw-hill-affects-135-million-accounts/ 🤖 Frontier AI Models Raise Concerns Over Offensive Capabilities What happened: Policymakers and researchers are raising concerns over increasingly powerful AI systems capable of autonomous vulnerability discovery and exploit chaining. Discussions are underway around leveraging these capabilities defensively, while limiting misuse as models become dramatically more capable. 🔗 https://datainnovation.org/2026/04/federal-government-should-partner-with-frontier-ai-labs-on-cybersecurity-defense/

You can access chatGPT 5.5 today. It looks like the most powerfull AI model for CyberSecurity yet. Even better than Claude Mythos. GPT-5.5 completed TLO end-to-end in 2 of 10 attempts, making it the second model to do so. Mythos Preview, the first model to solve TLO, did so in 3 of 10 attempts. https://www.aisi.gov.uk/blog/our-evaluation-of-openais-gpt-5-5-cyber-capabilities On the Expert-level tasks, GPT-5.5 achieves an average pass rate of 71.4% (±8.0%, 1 standard error of the mean), compared to 68.6% (±8.7%) for Mythos Preview, 52.4% (±9.8%) for GPT-5.4, and 48.6% (±10.0%) for Opus 4.7. On this measure, GPT-5.5 may be the strongest model we have tested.

https://youtu.be/E_zDr_3V-WU