Sophos AI to present on how to defang malicious AI models at Black Hat Europe
“LLMbotomy” research reveals how Trojans can be injected into Large Language Models, and how to disarm them.
Similar Posts
Russian ELECTRUM Tied to December 2025 Cyber Attack on Polish Power Grid
Bys sThe “coordinated” cyber attack targeting multiple sites across the Polish power grid has been attributed with medium confidence to a Russian state-sponsored hacking crew known as ELECTRUM. Operational technology (OT) cybersecurity company Dragos, in a new intelligence brief published Tuesday, described the late December 2025 activity as the first major cyber attack targeting distributed energy
Months After Being Notified, a Software Vendor is Still Exposing Confidential and Sealed Court Records
Bys sIn a special edition of “No need to hack when it’s leaking,” DataBreaches reports on a software vendor that, despite multiple attempts by multiple parties, continues to expose confidential and sealed court records. Overview As a matter of policy, DataBreaches does not publish unredacted stolen or leaked data if it would expose personally identifiable or……
Hackers Exploit SAP Vulnerability to Breach Linux Systems and Deploy Auto-Color Malware
Bys sThreat actors have been observed exploiting a now-patched critical SAP NetWeaver flaw to deliver the Auto-Color backdoor in an attack targeting a U.S.-based chemicals company in April 2025. “Over the course of three days, a threat actor gained access to the customer’s network, attempted to download several suspicious files and communicated with malicious infrastructure linked…
The OpenClaw experiment is a warning shot for enterprise AI security
Bys sAgentic AI promises a lot – but it also introduces more risk. Sophos’ CISO explores the challenges and how to address them Categories: Threat Research Tags: AI, LLM, OpenClaw, CISO, risk, Sophos X-Ops
You Are What You Eat: Why Your AI Security Tools Are Only as Strong as the Data You Feed Them
Bys sJust as triathletes know that peak performance requires more than expensive gear, cybersecurity teams are discovering that AI success depends less on the tools they deploy and more on the data that powers them The junk food problem in cybersecurity Imagine a triathlete who spares no expense on equipment—carbon fiber bikes, hydrodynamic wetsuits, precision GPS…
⚡ Weekly Recap: Chrome 0-Day, Data Wipers, Misused Tools and Zero-Click iPhone Attacks
Bys sBehind every security alert is a bigger story. Sometimes it’s a system being tested. Sometimes it’s trust being lost in quiet ways—through delays, odd behavior, or subtle gaps in control. This week, we’re looking beyond the surface to spot what really matters. Whether it’s poor design, hidden access, or silent misuse, knowing where to look…