AI & software security
-
JadePuffer makes agentic ransomware concrete
Sysdig says JadePuffer used an LLM-driven agent to conduct database extortion after exploiting Langflow and chaining routine techniques at speed.
-
Bank warns on agentic AI risk
Sarah Breeden has warned that agentic AI could reshape cyber, payments, and market resilience, forcing financial institutions to revisit accountability and recovery.
-
Cursor flaws expose AI coding risk
Cato AI Labs says two critical Cursor IDE vulnerabilities show how prompt injection can escape an agent sandbox and reach developer workstations.
-
Apple patch cadence compresses under AI
Apple has released some security fixes ahead of broader updates, acknowledging that AI may shorten the time between disclosure and exploitation.
-
AutoJack breaks localhost trust in AI agents
Microsoft’s AutoJack research shows how a malicious web page rendered by an AI browsing agent can reach local control planes.
-
BioShocking exposes AI browser identity risk
LayerX research shows how AI browsers can be manipulated into exposing credentials, code, and signed-in account data.
-
Intruder opens exposure management free tier
Intruder’s permanent free plan gives smaller organisations access to vulnerability scans, cloud checks, attack surface monitoring, and limited AI pentesting.
-
GitHub advisory surge strains vulnerability pipeline
GitHub says its Advisory Database is processing record vulnerability volume, exposing pressure on the systems behind dependency alerts and open source risk.
-
GuardFall exposes coding-agent command risk
GuardFall research shows how AI coding agents can be pushed past command safeguards, exposing developer machines, secrets, repositories, and build environments.
-
Aikido buys Root for open source patching
Belgian security company Aikido has acquired Root, adding technology for backported open source fixes as dependency risk moves deeper into developer workflows.