Skip to content
Breaking:

ServiceNow Calls for Context-Aware Kill Switches in Enterprise AI Governance

Bhakti Pitre, vice president of AI platform security product at ServiceNow, explains why halting malfunctioning AI agents requires balancing risk against underlying business workflows.

By The Company Wire3 min read
Share
ServiceNow — ServiceNow Calls for Context-Aware Kill Switches in Enterprise AI Governance
ServiceNow — ServiceNow Calls for Context-Aware Kill Switches in Enterprise AI Governance. Photo: SiliconANGLE.

As enterprises transition autonomous artificial intelligence agents into production workflows, the mechanics of containing malfunctioning systems have become a central governance challenge. Enterprise software provider ServiceNow Inc. is urging organizations to adopt a measured approach to agent containment, arguing that any response to an errant agent must weigh risk against the specific business processes it supports.

Speaking during a livestreamed interview on theCUBE at Okta Inc.’s Oktane conference, in coverage reported by SiliconANGLE (https://siliconangle.com/2026/09/28/servicenow-says-ai-agent-kill-switches-need-business-context-oktane/), Bhakti Pitre, vice president of AI platform security product at ServiceNow, warned against treating containment as a binary action. Pitre stated that an effective kill switch cannot function as a simple on-off switch across complex enterprise environments.

ServiceNow has spent the year positioning its platform as an operational control tower for enterprise AI. That effort includes post-deployment oversight through its expanded AI Control Tower, which structures governance across five phases: discover, observe, govern, secure, and measure. Under this framework, containment measures are intended to scale with the severity of the operational risk.

To illustrate proportional containment, Pitre contrasted failure scenarios. If an AI agent merely ceases delivering output, administrators may only need to pause the agent and conduct an investigation. In contrast, if an agent permitted to grant a 10% discount suffers a prompt injection attack and begins issuing 100% discounts, the incident warrants pulling the plug immediately.

Deciding when to intervene requires mapping an agent directly to the enterprise processes that rely on it. Pitre noted that ServiceNow maps these dependencies and integrates identity security technology from Veza Inc. to modify excessive permissions rather than unconditionally terminating functional workflows.

Visibility across distributed systems remains an industry hurdle, as no single vendor oversees every endpoint, network pathway, and gateway an agent traverses. To address these boundaries, ServiceNow is collaborating with Okta to secure agent session tokens and agent identities. Pitre emphasized that broader vendor coordination will be required across the software ecosystem to ensure enterprise AI can be safely deployed at scale.

Sources

  1. SiliconANGLE

Company: ServiceNow

Written by

The Company Wire

Newsroom · San Francisco

Inside the companies building what’s next. Reporting on startups, technology, funding and the people shaping them.