Google Kubernetes Engine stories
The move puts Agent Substrate, gVisor and OpenRL under neutral governance as companies race to secure and scale automated workloads.
The rollout aims to help companies patch AI-driven systems faster, as Google adds identity, perimeter and detection controls across its cloud security stack.
Enterprises can now run autonomous AI agents with tighter isolation and governance as Google Cloud expands controls across its own services and other environments.
The startup says more reliable data generation is keeping pre-training on schedule after throughput rose 4.07 times across its production clusters.
The open-source stack aims to cut latency and boost accelerator use for AI workloads that now arrive in bursts and compete with interactive requests.
Biotech researchers could gain tighter control over sensitive data and slower simulations as Google Cloud separates planning, computation and oversight.
The add-on lets security teams enforce runtime monitoring across large Kubernetes fleets without manually deploying agents or distributing API keys.
The update could cut engineering overhead for firms mixing web, backend and Android systems, as agents can now interoperate across five languages.
The service aims to cut storage overheads for firms running thousands of AI agents, while keeping workspaces isolated and quick to spin up.
Teams running AI agents at scale could save idle compute and tighten isolation as Google opens Agent Substrate on Kubernetes Engine.
Enterprises seeking AI-ready infrastructure may see Google Cloud's higher score as validation of its database, analytics and Kubernetes services.
Under heavy generation load, the smaller model kept scaling on TPU v6e while the 27B version hit a throughput plateau above 64 users.
The recognition bolsters Google's pitch to enterprises seeking AI-ready cloud systems with tighter data control as rivals vie for big workloads.
Enterprises facing AI demand spikes are being told to secure spare capacity and automate backups to avoid outages and overspending.
The experimental setup lets AI teams run untrusted model code in isolated Ray clusters on Google Kubernetes Engine without exposing host containers.
Customers face a 2029 deadline to test post-quantum controls as Google Cloud begins rolling out quantum-safe encryption across its services.
The preview could help GKE users enforce cluster-wide security without clashing with namespace rules, especially in multi-tenant environments.
The upgrades are aimed at easing bottlenecks for large AI and machine learning deployments, with clusters now scaling to 15,000 nodes.
Tests on an n2-standard-48 virtual machine showed Google Cloud's sandbox and orchestration tweaks could cut per-agent costs by up to 75%.
Retailers using Malachyte's Google Cloud system are seeing sales double as it personalises search and recommendations within 100 milliseconds.