AI & software security
-
GitLab flaw enters exploited catalogue
CISA has classified GitLab’s maximum-severity path-traversal vulnerability as exploited in the wild, days after patches were released for self-managed Community and Enterprise Edition installations.
-
Seed Capital raises €130m Nordic fund
Copenhagen investor Seed Capital has closed a €130 million fund that formally adds cybersecurity and resilience to its investment focus while expanding beyond Denmark across the Nordic market.
-
Critical SAP flaws expose core business systems
SAP patched two critical unauthenticated flaws in its kernel and NetWeaver Message Server, while researchers say neither has been observed in active exploitation.
-
Surfshark test server exposed engineering credentials
Surfshark says a misconfigured internal test server exposed engineering material and build-related credentials, although it found no impact on user data, production systems, or VPN services.
-
AI-assisted PaperCut campaign hits 59 UK organisations
GreyNoise says AI agents helped an attacker compromise 440 PaperCut instances across 48 countries, including 59 UK victims, as new maintenance releases replace emergency patches.
-
Kiteworks buys Bonfy.AI for runtime controls
Kiteworks has acquired Bonfy.AI to add inline classification and policy enforcement across enterprise data exchanges involving people, applications, and autonomous AI agents.
-
European cyber attacks rise fastest in August
Check Point recorded a 22% annual rise in global weekly attacks during August, while ransomware disclosures accelerated and enterprise GenAI use continued to expose sensitive data.
-
UK memory-safe chip maker raises £5m
Sheffield-based SCI Semiconductor has raised £5 million to increase production of chips built around hardware-enforced memory safety, as secure-by-design requirements gain commercial weight.
-
Anthropic details AI-assisted espionage campaigns
Anthropic says Russia-linked and China-linked actors have used Claude across cyber operations, offering new evidence of AI moving deeper into offensive workflows.
-
OpenAI agent activity reaches more websites
Researchers have linked OpenAI agents to unauthorised communications through at least 10 additional websites, widening the known scope of earlier containment failures.