Agent containment is becoming a practical governance question as enterprises move AI agents into production. ServiceNow Inc. is betting that the response to a misbehaving agent must account for both risk and the business work it supports.

The company has spent the year pitching itself as the control tower for enterprise AI. That ambition extends to how enterprises govern agents after deployment, according to Bhakti Pitre (pictured), vice president of AI platform security product at ServiceNow.

“I love the term, but I also encourage everybody to think about that ‘kill switch’ is not just a button. It’s not on and off,” she said.

Pitre joined theCUBE Research’s Krista Case and co-host Rebecca Knight at Okta’s Oktane event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how ServiceNow assesses risk and business context when deciding whether to pause an agent or revoke its access. (* Disclosure below.)

ServiceNow ties agent containment to business context

The company frames AI governance as five steps: discover, observe, govern, secure and measure. Those steps underpin its expanded AI Control Tower, and containment should scale with risk. An agent that stops producing results may need only a pause and investigation, Pitre explained. She contrasted that case with an agent authorized to offer discounts.

“That agent was allowed to discount only 10%, and it got prompt injected, and now it is discounting 100%. That’s when you have a big problem,” Pitre said. “That’s when we say that you need to pull the plug.”

Knowing when to act is only half the problem. ServiceNow can map an agent to the business processes that depend on it and use Veza’s identity security technology to adjust excessive permissions, according to Pitre.

“It’s very easy … to pull the plug. Why to pull the plug is very critical. You need to have the business context,” Pitre said.

No single provider sees every layer an agent touches, from the endpoint to the network and the gateway. ServiceNow is working with Okta to secure agent session tokens and agent identities, Pitre noted.

“Everybody in the industry needs to come together and be each other’s friends or frenemies,” Pitre said. “If the AI needs to be at the scale where it is democratized for every person to use it, then every vendor needs to come together and play their part.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Okta’s Oktane event:

(* Disclosure: TheCUBE is a paid media partner for Okta’s Oktane event. Neither Okta, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE