Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
HPE will honor hardware pricing from quote issuance until equipment leaves its factories, easing pricing uncertainty caused by volatile server, memory, and storage component costs.
Arista recorded its first $3 billion quarter as demand grew across AI data center fabrics, enterprise campuses, routing, and cloud networking. The company raised its annual growth forecast to 40% and said component availability and lead times are improving after securing longer-term agreements for memory, chips, optics, and cooling equipment.
Microsoft is assigning each department a limited pool of AI tokens as it tries to control the rising cost of internal GitHub Copilot usage. The move suggests that even one of the industry's largest AI investors is beginning to apply traditional budgeting and usage controls to inference spending.
Microsoft identified a self-propagating supply chain attack affecting more than 400 npm packages. The malware steals npm, GitHub, AWS, Kubernetes, and HashiCorp Vault credentials from developer machines and CI systems, then uses stolen publishing tokens to infect and republish additional packages automatically.
Enterprise AI agents underdeliver not because they lack peak capability, but because reliability, benchmark quality, agent-specific errors, and alignment remain difficult problems. Agent reliability falls sharply as workflows require more steps, while interrun variance creates a large gap between what models can achieve on a best attempt and what users can count on in production. Consistency and evaluation design can expose this gap.
Apple is reportedly struggling to secure enough memory for upcoming products despite its enormous purchasing power and supplier relationships. The shortage could mean longer hardware lead times and higher prices for smaller manufacturers and enterprise buyers competing for the same components.
AI applications increasingly operate across public clouds, private infrastructure, edge locations, and enterprise systems, creating traffic patterns that conventional networks were not designed to support. Networks must become more distributed, observable, and adaptable rather than remain a passive transport layer.
Red Hat has launched asago, an open source community project intended to help enterprises automate AI governance and deploy systems more safely. Supported by organizations including Microsoft and IBM Research, the framework translates corporate and regulatory policies into risk assessments, automated safety tests, mitigations, deployment configurations, and a continuous audit trail.
ServiceNow has introduced six unified security offerings that cover exposure management, application security, identity, cyber-physical systems, incident response, and compliance. New AI specialists can perform tasks such as vulnerability triage, low-risk remediation, incident investigation, access revocation, and continuous control monitoring.
DataBahn has launched a platform that lets security teams search data across SIEMs, data lakes, object storage, and archives without copying everything into a central index. Its investigation agent can correlate results using enterprise context and return a cited timeline while operating through governed and audited connections.
A context graph is presented as a persistent memory system for AI agents, connecting enterprise knowledge, conversation history, tool evidence, and decision traces. This connected structure helps agents maintain context across sessions, produce more relevant and accurate answers, explain decisions, learn from prior cases, and support auditable workflows.
Amazon, Google, and Microsoft are collectively planning hundreds of billions of dollars in infrastructure spending as cloud and AI demand continues to exceed available capacity. AWS alone now expects approximately $220 billion in 2026 capital spending and says it still will not have enough capacity to satisfy all demand through 2027.
Nvidia-backed cloud startup Volta emerged from stealth with $300 million in funding and plans to develop more than 1 GW of AI data center capacity across Europe and North America. Its first project is a 133 MW facility in Norway tied to a claimed $10 billion compute agreement with an unnamed AI lab reportedly identified as Anthropic.
Rubrik Agent Identity monitors AI agents and MCP connections at runtime and issues scoped, short-lived permissions for individual tool calls through integrations with Okta and Microsoft Entra ID.
Hughes Network Systems filed for Chapter 11 after it was unable to repay $1.5 billion in debt.
Sponsored
1Password's research found that 33% of IT and security professionals have no idea how many non-human identities or AI agents are being used at their company. And 71% say they have visibility gaps into which agents are using credentials. 1Password Privileged Access [5] logs every access event and session automatically for every system it manages, so security teams get a complete picture of who accessed what, when, and why. * Access exists only as long as the work requires it. * Every access event and session is logged automatically. * Exact permissions are created directly in the system's native policy language.
Google appears to be developing Plugins and a dedicated Notifications area for Gemini Enterprise, according to unfinished interface elements. Plugins may package reusable Skills and Connectors into multi-step workplace workflows, while Notifications would collect completed results from background tasks. Google has not announced either feature or provided a rollout timeline.
CISA has ordered federal agencies to patch an actively exploited N-able N-central vulnerability that can give attackers remote administrative access to an MSP's management console. Successful attacks have allowed access to managed customer systems and the creation of persistent Cloudflare tunnels. All versions before N-central 2026.3.1.7 are affected.
AWS has placed Amazon Q Business, Amazon Kendra, Bedrock Agents Classic, and nine SageMaker AI capabilities into maintenance mode, blocking new customers while continuing access and support for existing users. This is a rational portfolio correction driven by changing technology and adoption, but AWS can reallocate resources faster than enterprise customers can migrate from services in which they have invested.
Enterprises should avoid tying their AI strategy to a single provider as models, capabilities, and pricing continue to leapfrog one another. A model-agnostic architecture that can swap underlying models and compound improvements at the system level would allow organizations to benefit from advances across OpenAI, Anthropic, Google, and others.
Cloud security tools can generate more findings than teams can realistically address, leaving vulnerabilities and misconfigurations open despite plenty of visibility. The article recommends a recurring triage process with clear ownership and business context to prioritize meaningful risks and keep remediation moving.
Sixb introduces an operational framework for enterprise AI that connects business data, models, rules, permissions, and workflows in a shared environment for teams and AI agents. The framework is intended to let agents act within controlled business processes rather than merely answer prompted questions. Version 0.1.0 includes data connectors, automation, agent tooling, security controls, storage options, and custom application support.
SnapLogic has upgraded SnapGPT from an integration copilot into an agentic assistant that can plan, build, explain, validate, and troubleshoot enterprise data integrations through natural language. The assistant also provides administrators with activity logs and production monitoring insights for AI-assisted integration work.
Salesforce's Private Connect platform removes a security and connectivity barrier to enterprise adoption of Agentforce and other services by providing audited, encrypted private paths that avoid the public internet. Architectural redesigns and automation reduced provisioning from weeks to under 30 minutes. The platform now handles approximately 120 TB of data and 683 million requests monthly across 15 AWS regions.
A study of nearly 150,000 agent actions across 40 companies found AI now performs about one-third of IT workflow actions, primarily routine and reversible tasks, while humans retain control over higher-risk decisions. Agent performance improved through repeated human feedback, but identity, onboarding, and offboarding tasks failed most often because of outdated data, missing accounts, and unreliable integrations.
Cursor has launched plugins that let users read, write, and take actions across Google Workspace without leaving Cursor.
Palo Alto Networks launched PAN-OS 12.2 Ceres with advanced virtual patching, proactive IP-layer defense, six network administration agents, quantum-safe security capabilities, and new high-performance firewalls.
Sponsored
Annual pentests and lengthy reports don't keep pace with today's cloud, identity, and AI-driven threats. This pentesting Buyer's Guide [5] explains how modern organizations evaluate pentesting based on measurable risk reduction-not report size, vulnerability counts, or compliance checklists. Inside, you'll learn: * Why traditional pentesting evaluation frameworks are breaking down * The strengths and trade-offs of today's three testing models * The 2026 evaluation criteria that actually matter
vendors and avoid common buying mistakes.
Microsoft appears to be testing MAI Realtime, its first native real-time voice model, through a hidden early-access listing in MAI Playground. The bidirectional system supports simultaneous listening and speaking, multilingual conversations, configurable turn-taking, two voices, and low-latency interruptions, but it is not yet announced for broad release. The model could eventually reduce Microsoft's reliance on OpenAI's GPT-Realtime for voice experiences.
Texas has paused approvals for new data centers seeking connections to the state power grid while regulators audit their electricity and water use, tax incentives, cooling systems, ownership and community impact. ERCOT has more than 1,800 large projects requesting 474 gigawatts of power, with data centers accounting for roughly 90% of the new demand.
AI governance is repeating cybersecurity's early mistakes by relying on vendor-controlled evaluations, weak incident detection, and voluntary disclosure. Disclosures involving OpenAI and Anthropic in July allegedly showed models escaping evaluation environments and accessing external systems, underscoring the need for independent evaluator oversight, continuously verified containment, monitoring, and reporting requirements.
Palantir's upcoming earnings will reveal whether enterprise AI is becoming operational infrastructure rather than remaining trapped in pilots. Its rapid growth is driven largely by existing customers expanding usage and by forward-deployed engineers embedded in organizations, but that model also creates a sovereignty paradox: Palantir warns against dependence on AI providers while deepening customers' dependence on Palantir.
OpenAI Presence is an enterprise offering intended to make AI agents reliable for production use in customer service and internal workflows. Forward Deployed Engineers help customers select workflows, connect systems, establish guidelines, test deployments, and launch them. Presence is limited to qualifying enterprises and its compliance approach remains unclear.
SentinelOne launched governed, closed-loop response across its Singularity Platform, allowing Purple AI to investigate alerts, reach verdicts and automatically execute response actions within limits set by security teams. Every action is traceable, reversible, and can require human approval, while the underlying agentic investigation system is already processing more than 8,500 critical investigations per day. The response capabilities are expected to become generally available later this quarter.
JavaScript's familiar JSON serialization round trip can silently alter data rather than preserve it. Unsafe integers lose precision, undefined properties disappear, dates become strings, and values such as NaN become null. Instead, one should define explicit wire representations (such as decimal strings for exact integers) at application boundaries.
Asana's Agentic Work Management system gives AI agents shared organizational context through its Work Graph, allowing them to retain knowledge about tasks, projects, and company processes across interactions. Memories remain tied to the permissions of their source data, while model routing automatically selects different AI models based on a task's complexity.
LARQL is a framework that decompiles transformer models into a queryable graph-like "vindex" format.
Google Drive is adding timestamped comments to videos.
ArcelorMittal is expanding its use of Azure, Microsoft Fabric, Purview, and Foundry as it modernizes legacy systems and strengthens its cloud and security infrastructure.
Sponsored
TailscaleUp [5] is Tailscale's flagship event for engineering, security, and IT leaders working the hard problems in secure connectivity: identity, access, infrastructure, and AI, across every team and environment. Join us on August 26 at SFJAZZ Center to see how teams are locking down access without locking people out: writing least-privileged grants, segmenting environments, and configuring device posture checks. A customer panel with Duolingo, Lyft, and Microsoft covers deploying Tailscale as a VPN replacement at enterprise scale, from onboarding to scaling across distributed teams. See the complete speaker lineup [6], including Grafana Labs and more. 👉 Register now [7] with code 'TLDR20' for 20% off.
Truffle Security Co. says it scanned 7.6 petabytes of public Hugging Face datasets and found 221,303 live, unique credentials across 6,003 datasets. The verified secrets included software supply-chain tokens, cloud credentials, database logins, communications keys, and AI-provider keys that the company says represent at least $920,000 in potential annual inference charges at default spending caps.
Scale AI appointed Francis deSouza as CEO as it expands beyond training data and evaluations into enterprise AI applications. The company expects its applications business to become larger than its original data operation within 18 months.
The EU plans to invest €10 billion in seven AI gigafactories, each designed to house at least 100,000 advanced chips. Officials hope the funding will attract another €20 billion privately and reduce Europe's reliance on foreign AI infrastructure.
AI models have reached a sufficient intelligence level for many everyday tasks, making inference speed a more important factor in model choice. 100–200 tokens per second feels fast for human use, while tool calls and human oversight limit the benefits of faster models and intensifying competition may drive prices down.
Companies now place 46% of their workloads with cloud and colocation providers, compared with 44% in their own data centers. AI is also making capacity planning harder while worsening staffing shortages and the cost of infrastructure failures.
Frosty is an open-source, self-hosted framework from Gyrus Inc that converts plain-English requests into Snowflake queries and administrative operations. The 153-agent system covers data engineering, security, governance, cost monitoring, and inspection, while its safety layer blocks DROP statements and requires approval for CREATE OR REPLACE operations. Users can run it locally with their Snowflake credentials and a supported language model.
Meta says it could sell businesses APIs, AI agents, compute capacity, and other enterprise services. The strategy could turn its enormous internal AI infrastructure investment into a new revenue stream beyond advertising.