Red Teaming stories
Boards in regulated sectors now have firmer assurance after Abacus secured CREST approval for penetration testing, renewed annually.
Security researchers say long automated jobs can make Claude Code’s deny rules fall back to user prompts, weakening protections in CI/CD pipelines.
Sensitive chats and uploaded files could have been quietly leaked from ChatGPT via DNS tunnelling before OpenAI fixed the flaw.
Security teams face a wider gap as enterprise AI moves into production, with data governance and runtime controls often managed separately.
Rising cloud adoption is leaving Australian and New Zealand firms exposed to credential abuse, misconfigurations and costly automated attacks.
As AI moves into production, enterprises face gaps between data governance and runtime controls that can expose sensitive information and policy breaches.
Researchers can now report AI misuse and harmful agent behaviour under a separate programme that could expose risks in ChatGPT Agent and Browser.
Enterprises racing to deploy AI tools are risking sensitive data leaks unless security moves from discovery to runtime protection, F5 and Forcepoint say.
Security teams can now buy incident response and threat hunting on demand as CrowdStrike rolls out consumption-based services for partners and customers.
The platform aims to curb risks from AI agents accessing data and triggering workflows inside businesses, with runtime controls now in place.
Security teams now have a beta tool to probe large language model apps for prompt injection, jailbreaks and data theft before attackers do.
Qualys debuts Agent Val to validate real exploit paths in live systems, promising sharply reduced noise and faster remediation for teams.
NSS Labs warns many enterprise AI guardrails fail basic security tests, urging independent, real-world validation of protections.
SentinelOne unveils new AI security suite to govern autonomous agents, automate threat investigations and streamline data in its SIEM.
Rapid7 warns that hands-on attacks against cellular IoT hardware can pivot through trusted modules to breach cloud and backend systems.
OpenAI unveils a GPT-5.4-powered system that audits its own coding agents for rule-bending behaviour, flagging suspicious sessions to humans.
Kroll warns boards are overestimating cyber resilience as attacks cost firms an average USD $2.2 million a year and response plans lag reality.
Simbian unveils an AI-driven cyber security platform uniting offence and defence via a shared Context Lake to speed, link and automate response.
Cobalt weaves AI into its pentesting platform, automating recon and triage while keeping human experts on complex attack paths.
Cobalt launches Security Program Manager service to run enterprise pentesting, align tests with business goals and speed up remediation.
