We’ve already established that modern work runs inside the browser. The more uncomfortable truth is where breaches actually succeed.
They don’t succeed at the firewall. They don’t succeed at the inbox.
They succeed inside an authenticated session, operating under a valid identity, inside the workflows your organization relies on every day.
We tend to describe breaches in terms of how they begin. A phishing email bypassed filtering. A credential was stolen. A device was compromised. Those events matter. They aren’t where the real damage occurs.
The decisive moment comes later, when a valid identity is used in a trusted session to perform actions the system implicitly allows.
The trusted browser session is the new blast radius
Modern enterprise work now runs predominantly inside browser-based applications. As discussed previously, workforce research estimates that roughly 85 percent of enterprise work takes place in the browser. That same environment has become a primary channel for browser-based attacks.
What makes this environment particularly valuable to attackers isn’t a lack of controls at login. It’s the inheritance of trust that follows authentication.
Once a user is logged in, the application behaves as designed. Features function normally. Data flows through legitimate interfaces. From a system perspective, there is no obvious boundary being crossed.
CrowdStrike’s Global Threat Report has highlighted how rapidly attackers now move once access is obtained, with breakout times often measured in minutes. When activity unfolds inside an already authenticated browser session, that speed compresses the window between access and impact even further.
The session isn’t a side channel. It’s the operating environment.
AI increases the velocity inside the session
AI hasn’t changed the fundamentals of how breaches succeed. It’s changed the tempo.
Recent research from Microsoft on AI-assisted phishing shows significantly higher engagement rates compared to traditional campaigns. The messages are more context-aware and more convincing. But the real acceleration happens after the click.
AI tools live in the browser. Sensitive information is pasted into prompts. Files are uploaded directly through web interfaces. In one 2025 survey, 38% of respondents admitted to accessing generative AI tools routinely. Now imagine how many of your employees are using browser based AI tools in 2026. Activities that once moved through slower, structured processes now move through a single interaction.
The effect isn’t a new category of attack. It’s the compression of time between action and consequence inside a trusted session.
The flaw in the perimeter narrative
We still frame breaches as failures of entry controls. A malicious email slipped through. A credential was reused. An access token was exposed. In practice, the impact materializes when someone inside an authenticated session performs an action that is entirely consistent with their permissions. Exporting customer data from a SaaS application. Changing vendor bank details. Entering credentials into an external tool.
These are legitimate capabilities used in unintended ways. Authentication validates the identity, not the intent. If the last meaningful control point is login, then everything that happens after login inherits trust by default.
In a browser-dependent, AI-accelerated environment, that design choice becomes the most consequential weakness in the stack.
The question that matters
If breaches consistently succeed inside trusted sessions, what governs behavior after authentication?
If valid identities can be used to move sensitive data through sanctioned applications, where does control actually live?
The browser is now the runtime for modern work. Attackers understand that.
Security architecture needs to reflect it.
If it doesn’t, then the most important phase of a breach is unfolding in the one place we still assume is safe.