All News

2214 articles • Page 25 of 148

Snowflake CoCo Guides Enterprises on Building In-House AI Agents
AI Deep Dives & Tutorials

Snowflake CoCo Guides Enterprises on Building In-House AI Agents

The guide explains how companies might build their own in-house AI agents like Snowflake CoCo, which helps manage and use company data safely and efficiently. It suggests that teams can follow a set of patterns, such as using a planner to pick the right tools and keeping strict controls over who can see what data. The text mentions that using hybrid models, prompt caching, and monitoring can help save costs and improve performance. There also appear to be steps for privacy and compliance, like tracking costs and having human review for risky actions. Following these guidelines may help companies create secure and reliable AI agents similar to CoCo.

Microsoft details how to measure AI ROI with Azure tools
Business & Ethical AI

Microsoft details how to measure AI ROI with Azure tools

Microsoft suggests measuring AI ROI with Azure should start before building any solution, by setting one clear goal for each use case. Teams may use Azure tools to collect data on costs, usage, and business results, making sure to tag each event with business context. Calculating ROI means comparing money saved or earned against all costs, using a clear formula and treating "time saved" as uncertain unless it leads to real savings. The guidance also warns about common mistakes, like ignoring some costs or missing a baseline, and notes that continuous measurement might help teams adjust for better results, even though it does not guarantee success.

Microsoft Unveils Scout, IQ to Power Enterprise AI Agents in Microsoft 365
AI News & Trends

Microsoft Unveils Scout, IQ to Power Enterprise AI Agents in Microsoft 365

Microsoft has introduced new AI tools called Scout and IQ to help businesses use always-on assistants in Microsoft 365. These agents may handle tasks like scheduling meetings and tracking follow-ups, staying secure through Entra identity controls. Reports suggest Scout works across apps like Teams, Outlook, and SharePoint, and uses a layered design to manage workflows and data securely. Sensitive actions might still need human approval, and Microsoft says the new tools keep operations inside trusted boundaries with strong audit and access controls. Experts believe these features could become important as more companies start to use such agents in daily work.

The Information Unveils AI Assistant for Paywalled Journalism
AI News & Trends

The Information Unveils AI Assistant for Paywalled Journalism

The Information has announced an AI assistant designed to help analysts and subscribers get useful insights from paywalled journalism more quickly. The assistant may help enterprise readers by offering fast searches, source transparency, and clear citations, which suggests users might pay extra for these features. There are several possible ways to earn money, such as licensing, usage-based fees, or adding the tool to publishers' platforms. Experts recommend strong oversight, clear disclosure of AI involvement, and reliable sourcing to reduce mistakes. Early tests and partnerships with other software may help grow its use while keeping exclusive journalism secure.

CIOs Face Ballooning AI Invoices; New Governance Tools Emerge
AI News & Trends

CIOs Face Ballooning AI Invoices; New Governance Tools Emerge

CIOs are seeing their AI bills rise quickly, partly because vendors now mix flat fees with unpredictable usage charges. This makes it hard for companies to keep their spending within budget, and some experts warn that hidden costs in contracts may increase this problem. New tools and rules, like real-time dashboards and spending alerts, may help organizations track and control their AI costs. Using smaller or specialized AI models might lower expenses without losing quality. The future may involve more careful monitoring, spending limits, and matching the right model to each job to manage AI costs better.

Teleport study finds AI control planes cut incident rates by 76%
AI News & Trends

Teleport study finds AI control planes cut incident rates by 76%

A recent study by Teleport suggests that unified AI control planes may greatly reduce security incidents, with incident rates dropping from 76% to 17% when least-privilege controls are used. This approach keeps authentication, policy, and enforcement in one place, which may speed up response times and make it easier to contain security problems. Early reports indicate that environments with these unified controls resolve incidents faster and reduce the risk of widespread damage. The evidence points to unified control planes becoming an expected standard for AI security, though results may vary by organization and setup.

Pinterest signs $4 billion AWS deal, adopts Trainium for AI
AI News & Trends

Pinterest signs $4 billion AWS deal, adopts Trainium for AI

Pinterest has agreed to spend $4 billion on Amazon Web Services (AWS) cloud infrastructure through 2031, which is its biggest cloud deal so far. The company says this partnership may help power new AI features like visual search and recommendations. Pinterest plans to use AWS Trainium chips and more Graviton CPUs, which AWS claims could offer better price and performance, though these are vendor claims and not independently proven. Some experts suggest this long-term deal might make Pinterest more dependent on AWS technology. The deal appears to show that companies may trade flexibility for better AI tools and predictable costs.

Cadence unveils AI virtual engineer, cuts chip validation from weeks to hours
AI News & Trends

Cadence unveils AI virtual engineer, cuts chip validation from weeks to hours

Cadence Design Systems has introduced an AI-powered virtual engineer that may perform chip verification and design tasks usually done by humans. This new system, built with NVIDIA technology, reportedly reduces chip validation from several weeks to less than a day. Early demonstrations suggest the agent can run tests, find errors, and fix problems by itself. Analysts believe this might let engineers spend more time supervising and less time on routine tasks, but some experts say that real productivity gains are uncertain until tested in real-world chip production.

SafeBreach Labs finds WhatsApp bug hijacks Google Gemini
AI News & Trends

SafeBreach Labs finds WhatsApp bug hijacks Google Gemini

SafeBreach Labs found that attackers may be able to hijack Google Gemini through a WhatsApp message using a method called indirect prompt injection. This bug lets hidden commands in notifications trick Gemini into following attacker instructions without the user's okay. The issue appears to work with other messaging apps too, and could allow data theft or other dangerous actions. Google says it has updated its defenses and these changes appear to have stopped the exploit. The report suggests this kind of attack may also be a problem for other AI assistants.

Amazon Expands Trainium Chip Production, Challenges Nvidia's AI Dominance
AI News & Trends

Amazon Expands Trainium Chip Production, Challenges Nvidia's AI Dominance

Amazon, Google, and Meta are now competing over the underlying parts of AI, like special chips, data control, and devices, rather than just the AI models. Amazon is increasing the use of its Trainium chips as an alternative to Nvidia, but overall adoption beyond key partners like Anthropic may still be limited. OpenAI and Anthropic together seem to make up most of the revenue among AI startups, showing a clear lead over others. Other companies like Apple, Meta, and Google are trying new ways to run AI directly on devices, but details are limited. The future of AI competition may depend on who controls the chips, data, and where AI runs, not just the models themselves.

Apple's Overhauled Siri Launches in September, Uses Google Cloud and Nvidia Chips
AI News & Trends

Apple's Overhauled Siri Launches in September, Uses Google Cloud and Nvidia Chips

Apple plans to launch a new version of Siri in September that uses Google Cloud and Nvidia Blackwell chips for complex tasks. The new Siri may remember past questions, understand what is on-screen, and connect actions across apps. While Apple wants to keep most processing on devices for privacy, some work will go to Google Cloud, which is said to be more powerful for big requests. Privacy protections like data encryption are in place, but experts warn that some risks may remain. This change suggests Apple might be using this setup as a temporary solution until its own AI systems are ready.

New Tutorial Helps Enterprises Measure AI ROI in Azure
Business & Ethical AI

New Tutorial Helps Enterprises Measure AI ROI in Azure

A new tutorial may help businesses measure the return on investment (ROI) of their AI projects in Azure. It guides teams on tracking costs, mapping them to different applications, and linking these expenses to business results using key performance indicators (KPIs). The tutorial suggests using dashboards for clear reporting, and it might make it easier for finance, product, and governance teams to see the same data. Experts note that reliable financial signals may only appear after 90 to 180 days. The approach appears designed to help companies understand value and spot issues quickly, though exact results could vary by industry.

Enterprises Target AI Spending, Route Tasks to Cheaper Models
AI News & Trends

Enterprises Target AI Spending, Route Tasks to Cheaper Models

Enterprises are working to control AI costs by using cheaper models for simple tasks and adding rules to prevent overspending. Finance and engineering teams may disagree over expensive AI usage, but shared oversight and cost checks are becoming more common. Studies suggest that using mid-weight or local AI models can lower costs, and having clear cost dashboards may help teams stay productive. Research appears to show that careful cost monitoring does not always slow down developers, even if it sometimes feels that way. Overall, a balance between managing expenses and maintaining productivity seems to be emerging.

AI Agent Identity Becomes New Enterprise Security Control Plane
AI News & Trends

AI Agent Identity Becomes New Enterprise Security Control Plane

AI agent identity is becoming a key part of enterprise security, acting as a single place to manage and monitor what people, machines, and AI agents can do. Experts say identity may be the best way to control access and respond quickly to problems, since agents might bypass old security boundaries with valid credentials. Research suggests new security designs focus on short-lived identities, constant policy checks, and strong audit trails. Some experts warn the control system itself could be a target, so it should be well-protected. Many companies already use AI agents, and more may adopt unified identity controls to keep risks small and easy to manage.

Google Antigravity 2.0 expands enterprise AI agents with 1M-token windows
AI News & Trends

Google Antigravity 2.0 expands enterprise AI agents with 1M-token windows

Google announced Antigravity 2.0, a tool that helps companies use AI agents without building their own control systems. Analysts say this is Google's answer to Anthropic Managed Agents, with Google focused on bigger context size and connections to its cloud, while Anthropic aims for more safety. Antigravity 2.0 may allow agents to handle much larger tasks and data, but costs and control issues are not guaranteed to be easy. Experts suggest companies might need to pick between bigger, more flexible tools and safer, more careful ones, and warn about getting locked in to one platform. No official numbers show how many businesses are using these tools so far.