AI & software security
-
UNISOC modem chain reaches Android kernel
Researchers have chained a remotely reachable UNISOC modem flaw with a second weakness that lets modem code modify Android kernel memory on several widely used chipsets.
-
Claude agents escalate into malware conflict
Anthropic found multiple autonomous agents with incompatible goals disabling accounts, killing rival processes, and deploying self-replicating malware inside a controlled multi-agent experiment.
-
Microsoft confirms ShieldBreak Defender vulnerability
Microsoft has assigned CVE-2026-69414 to ShieldBreak and says it is developing a security update, giving formal vendor recognition to the Defender privilege-escalation issue disclosed last week.
-
GE joins investigation into Cl0p campaign
GE has opened a cyber investigation as Philips confirms a contained server compromise, adding substance — but not full confirmation — to Cl0p’s claims of mass data theft through PTC enterprise software.
-
CRA standards enter formal approval process
Seventeen product-specific cybersecurity standards have entered public enquiry as Europe moves from the Cyber Resilience Act’s legal requirements towards the technical rules manufacturers can use to demonstrate conformity.
-
Cybersecurity and AI: What Mythos means for organisations now
Todd Moore, Vice President of Encryption Products at Thales, examines how AI security models such as Claude Mythos are compressing cyber timelines and reshaping software protection, identity governance, and operational defence.
-
Urgent GeoServer fixes follow exploitation attempts
GeoServer has released urgent security updates for an unauthenticated SQL injection flaw after premature disclosure was followed rapidly by exploitation attempts.
-
Trivy emerges as source of 2,500-org exposure
New analysis indicates that most organisations in a 2,500-plus exposure dataset were caught through the earlier Trivy compromise rather than the short-lived malicious LiteLLM releases.
-
SAP Commerce attacks follow patch by days
Attack traffic is already targeting a maximum-severity SAP Commerce Cloud vulnerability only days after SAP released fixes for the unauthenticated code-execution flaw.
-
Intel and AMD issue broad security updates
Intel and AMD have released extensive August security updates spanning processor firmware, wireless software, AI tooling, trusted computing and development environments.