I’ve been watching how security, privacy, and compliance are featuring in the latest big tech earnings reports, and this week’s investor calls from Microsoft, ServiceNow, and Alphabet offer a useful snapshot of where enterprise AI risk is heading.
All three vendors are tackling the challenge of how to govern increasingly capable AI agents. Their comments offered a clear view of the challenge facing enterprise buyers in finding the balance between giving AI agents enough access to do useful work without allowing them to expose customer data or take unauthorized actions.
The common thread from the CX perspective is control: who can access customer information, what automated systems can do with it and how quickly businesses can respond when something goes wrong.
Cryptocurrency exchange Coinbase’s disclosure this week that an incident at analytics provider Amplitude exposed contact information belonging to a small subset of its customers highlights how third-party suppliers can introduce customer-facing risks even when a company’s own systems remain secure.
Microsoft Warns Enterprise AI Needs Stronger Security Controls
Microsoft’s results for the first quarter of its 2027 financial year arrived alongside discussions about how enterprises can manage the security implications of AI agents.
“Our approach is to separate intelligence from authority, enforcing identity, permissions, monitoring and containment outside the model itself,” said Microsoft CEO Satya Nadella.
This distinction is important as agents gain the ability to retrieve information, use enterprise tools and execute workflows with limited human intervention.
The vendor highlighted the role of Microsoft Purview in auditing agent activity, indicating that the platform has audited more than 7.6 billion agent activities, with activity increasing fivefold quarter over quarter.
The wider challenge is determining whether an agent has permission to perform a particular action, rather than simply whether it can technically access a system. An agent connected to a CRM platform, for example, might need permission to retrieve customer information but should not automatically receive authority to change account details, issue refunds or disclose sensitive records.
These controls become especially important when agents operate across customer service, sales and back-office systems. Excessive permissions could allow a compromised or misdirected agent to expose personal information or make changes that are difficult to reverse.
The security question for enterprise buyers is how effectively identity management, monitoring and containment controls can keep pace with the growing number of agents operating across their environments.
ServiceNow: Agent Governance Becomes an Enterprise Priority
ServiceNow’s third-quarter results put AI governance and security alongside its expanding enterprise AI business.
The IT service management vendor’s security strategy focuses on giving organizations visibility into their AI estate and control over what agents can do. ServiceNow’s AI Passport solution, announced earlier in October, is designed to verify the authorized scope of first- and third-party agents at runtime. The company said its Super Intelligence Control Tower can help organizations discover agents, map access permissions, monitor activity and enforce restrictions.
The approach is intended to address a growing problem for enterprises as they introduce agents through different platforms, departments and third-party applications, leaving security teams without a complete picture of what is running or which information it can reach.
Bill McDermott, Chairman and CEO at ServiceNow, said:
“Businesses in every industry need a harness for enterprise AI. This is more than orchestration, governance, or security alone. It’s the nexus of all three categories coming together. We’re now in the rules and rails phase of enterprise intelligence.”
ServiceNow also announced a Control Tower Alliance involving companies including Anthropic, AWS, Cisco, NVIDIA, Okta, OpenAI and others, aimed at supporting governed enterprise AI adoption.
The practical concern for CX operations is whether agents can be stopped or restricted when they behave unexpectedly. An agent working across a contact center, customer database and order-management system could otherwise create problems across several workflows before an incident is detected.
The challenge for buyers will be ensuring that governance works across different vendors and infrastructure, rather than relying on controls available within a single platform.
Alphabet: Google Raises the Bar for Agent Security
Alphabet’s earnings call also addressed the security requirements emerging as AI systems take on more complex tasks.
As Google and Alphabet CEO Sundar Pichai said:
