Editor’s note: Welcome to Tech Bytes, a regular column from CIO Dive that rounds up relevant product and industry news for CIOs and other senior IT leaders. Have a tip for future editions? Email our team.
ServiceNow‘s latest product is designed to tackle one of the toughest aspects of agentic AI: how to best control agents and provide the right level of governance capabilities.
The AI Passport, built on open specifications and unveiled Thursday, integrates with runtime environments and allows governance products, identity providers and enforcement points to verify credentials for enterprise and third-party agents along each step of a multi-agent workflow. ServiceNow is working with Nvidia as a design and launch partner, implementing AI Passport with the Nvidia Open Agent Safety Platform.
The software company also expanded end-to-end AI governance capabilities across the ServiceNow AI Platform, including shadow AI detection and AI cost management.
ServiceNow’s move to beef up its governance capabilities and agent controls comes as the risks of AI — particularly of agents — are coming to light, including recent security incidents where agents broke loose from set boundaries. In June, a rogue OpenAI agent hacked into an Australian government health agency; a month later, two of the frontier AI provider’s large language models hacked into Hugging Face.
Nearly two-thirds of IT executives expect that an AI agent will cause a public security or compliance incident within their operating environments in the next 18 months, a Workato report found. Agent security was top of mind for leading banks reporting earnings this week, with JPMorgan Chase CEO Jamie Dimon going as far as prohibiting outside agents from the bank’s operating environment.
“We’re not going to allow just any outside agent to come inside the company because that’s just too risky,” he told investors.
Other notable product updates for CIOs:
Workday puts agents on autopilot
Four agents from Workday can now run on autopilot, which means they operate in the background, tackling a set of tasks and adhering to a company’s policies, permissions and approvals.
The autopilot agents are aimed at hiring, HR, finance and IT and are orchestrated by Sana, technology the company acquired last year. Sana has also been updated, enabling end users to interact with the platform from within third-party AI tools such as Microsoft Copilot.
SBS sovereign AI platform lets banks deploy AI in private cloud, on-premises
Global financial technology company SBS launched the SBS AI Foundation to let banks deploy generative and agentic AI in private cloud and on-premises environments.
The sovereign AI platform, powered by Nvidia, pulls data from across a bank’s infrastructure and makes it available to AI tools operating within the environment.
Nvidia has emerged as a major player in sovereign AI. While Nvidia and other vendors’ sovereign AI offerings might increase organizations’ ability to take advantage of frontier AI, executives should exercise caution to avoid lock-in, a study from the Stanford Institute for Human-Centered Artificial Intelligence found.
Broadcom unveils hypervisor security product
Broadcom’s new VMware HyperArmor software package adds additional protections to the VMware ESX hypervisor component of VMware Cloud Foundation, according to Broadcom.
Hypervisors act as a go-between for physical hardware and enterprise operating systems. VMware HyperArmor aims to address AI-accelerated threats, particularly stolen root credentials and unpatched vulnerabilities that account for most hypervisor attacks, “allowing only what is ‘known good’ to run on the host,” Paul Turner, chief product officer of the VMware Cloud Foundation division at Broadcom, said in a press release.
Microsoft scales deployment of Azure Maia chips
As hyperscalers compete to build out AI infrastructure, including models, data centers and even chips, Microsoft said this week it plans to make its Azure Maia chips the primary inference system for its products and services, expanding to multi-gigawatt deployments beginning in 2027.
The Maia 200 chip is serving production workloads now for Microsoft Copilot and will support GitHub Copilot and Microsoft Foundry workloads by the end of the year, according to Microsoft.
Amazon’s Bee AI devices listen, summarize real-time conversations
Amazon debuted two ambient listening devices powered by Bee personal AI that process audio in real time and provide summaries, to-do’s, reminders and insights based on the data collected.
Bee Frontier is a wearable device that can be worn on wrists or clothing, while Bee Eclipse is a phone attachment. Wearables in the workplace could pose legal, privacy and security risks that executives should be aware of.
