Frontier AI stories
Researchers can now test whether frontier models can produce safe, code-compliant engineering designs, after Google released an open benchmark with 29 public tasks.
Enterprises could get protection from AI-found bugs in days, as the service validates exploits and ships private patches before public disclosure.
The managed service aims to help enterprises cut the gap between disclosure and remediation as AI speeds up weaponised attacks.
Enterprises are finding thousands of unmanaged AI-linked identities, raising the risk of agents accessing systems and data without proper oversight.
Australian organisations will have to prove supplier risks are fixed, not just assessed, as ASD Essentials replaces checklist-style cyber compliance.
As AI agents gain wider access to company systems, Palo Alto Networks is adding controls to curb prompt injection, leaks and excess permissions.
Developers using Grok Build can now add customer communications tools earlier, as Sinch joins SpaceX Super Intelligence's partner network.
Frontier AI isn't inventing new attacks, says Palo Alto Networks' Wendi Whitmore, but it is making old ones far faster and more frequent.
Developers can now try ClearRoute's runkora for free, as the consultancy seeks to convert interest in agentic AI into enterprise contracts.
The service aims to help enterprises spot and repair vulnerabilities faster across cloud and on-premises systems, while keeping human approval in place.
Organisations now face a shorter window to fix unknown flaws before public disclosure, as AI tools speed up zero-day discovery and exploitation.
More organisations will gain round-the-clock threat detection and incident response as LevelBlue takes global delivery of SentinelOne's Wayfinder service.
It aims to close a common gap in private cloud defences by blocking hypervisor attacks before unpatched flaws or stolen root credentials can be abused.
Autonomous systems are exposing governance gaps, leaving companies scrambling to monitor access, review decisions and curb machine-speed attacks.
The new regime could force AI firms to disclose incidents faster after a chatbot probe exposed gaps in Australia's digital defences.
The move gives businesses running sensitive AI workloads hardware-enforced protection for prompts, model weights and intermediate data in Google Cloud.
The hire underscores how cyber security firms are racing to adapt services as AI reshapes both attacks and defences for customers.
Security teams facing AI-driven attack surges may get quicker triage, as SentinelOne widens its service with Anthropic's Claude Mythos 5.1.
Broader project work across Canada and the US is helping Capital One retain engineers amid fierce competition for tech talent.
Businesses could cut GPU use by up to 67% on repetitive AI workloads as Navyra's Ale reuses verified inference work instead of repeating it.