Akamai report highlights how securing agentic AI enterprise needs behavioral governance

Enterprise AI is moving from tools that assist employees to agents that can act across APIs, applications and business systems. That shift is creating a security problem that traditional identity and access controls may not fully address.

Akamai’s latest State of the Internet Security report argues that securing agentic AI enterprise environments now requires a stronger focus on behavioural governance. The concern is no longer only who has access, but what autonomous systems are doing, how they behave and whether their actions can be verified.

Nonhuman identities change the security equation

As AI agents take on more tasks, enterprises are dealing with a growing population of nonhuman entities. Akamai identifies this as a shift from managing human identities towards governing autonomous systems.

The challenge becomes harder because AI models can discover and chain system weaknesses faster than traditional human-led patch cycles can respond. Akamai therefore points to real-time edge mitigation alongside backend patching.

The Model Context Protocol adds another visibility concern. While MCP allows AI agents to interact with multiple software systems, Akamai says MCP exposure ranks last among current CISO security priorities, even as security leaders expect rogue AI agents to become a major cyberthreat by 2030.

Agentic AI security also reaches the browser

The browser is becoming another important part of agentic AI security. More than 40% of enterprise users have installed AI-powered browser extensions, according to the report, while 25% of those extensions changed permissions within 12 months.

Akamai also says these extensions are 60% more likely to have known CVEs than standard extensions. This makes browser visibility and behavioural controls increasingly important as employees adopt AI tools outside traditional enterprise controls.

Chatbot use creates another exposure. More than 6% of enterprise AI chatbot conversations contain sensitive corporate information, while 47% of interactions take place through unmonitored personal accounts.

CISOs need governance that matches autonomy

Akamai’s recommendations point towards behavioural governance rather than simply adding more access controls. Security teams are urged to apply adaptive edge protections, secure browser-based AI use, and protect corporate information as AI-driven discovery grows.

The report also recommends matching an agent’s level of autonomy to how easily its actions can be checked and reversed. High-stakes actions should retain human oversight.

Read More: 

Googlebook Intel Core Ultra Series 3 launch brings AI laptops

Microsoft Data Centre expansion signals a regional shift

Commvault SHIFT 2026 to focus on AI, cyber recovery and resilience

Lalit Choudhary Takes Charge as ISODA President