Cloudbrink boosts AI security for hybrid enterprises

Cloudbrink has announced new security features aimed at managing enterprise use of AI agents and browser-based AI services alongside user access controls on its secure connectivity platform.

The company said the additions sit with its zero trust network access offering and focus on policy, visibility and traffic classification for AI-related activity in hybrid work environments.

AI adoption across enterprises has widened the range of tools and connection paths used for sensitive work. Some organisations use hosted AI services through a web browser, while others develop AI agents, plugins or run models locally. That mix can create multiple routes for data to move between users, applications and external services.

Cloudbrink cited research that indicates broad enterprise uptake of AI, alongside security concerns linked to varied platforms, different protocols and limited standardisation.

"AI is complicating the threat map. Enterprises are using AI in multiple ways across disjointed paths and every path needs to be secured," said Prakash Mana, CEO, Cloudbrink. "This becomes even more complex for companies with hybrid workforces. Last year was the year that companies dabbled in AI. In 2026 more enterprises are using AI for serious business and that requires security, scalability, and speed. That's what Cloudbrink is providing for AI in the enterprise."

Safe AI

Cloudbrink said the update expands what it calls its Safe AI features. The company described the work as an effort to align controls for AI agents and online AI services with existing controls for users and applications.

One of the new elements is Safe AI BrinkAgent. Cloudbrink said BrinkAgent sits as a component of its platform and can recognise traffic associated with AI agents and browser-based AI services. The company said the agent can also identify cases where sensitive data may be leaked. Cloudbrink said BrinkAgent acts based on security policies set by the customer and aligned to data protection and compliance requirements.

Cloudbrink also described a built-in database of definitions for AI agents and online AI services. The company said the database recognises protocols and platforms used by a range of AI services. It said the definitions update on an ongoing basis as new agent types and protocols emerge.

Custom definitions

For organisations that build internal AI agents or use specialist industry tools, Cloudbrink said customers can add custom AI agent definitions to the database.

The company also highlighted what it called unified policy and visibility. Cloudbrink said customers can manage security for hybrid workforce users and AI agents and services from the same console. It said the console provides visibility across users, applications, AI agents and services, as well as traffic details.

The company positioned the move as a response to the operational reality of enterprises deploying AI across different departments and workflows. AI agents can operate with varying degrees of autonomy. Browser-based AI services can sit outside typical application access patterns. Both can introduce new data flows that security teams must monitor and control.

Customer use

Cloudbrink said customers have already used elements of its AI platform since it introduced the offering last year.

One customer, Supervity AI, said security and trust issues remain central to broader adoption of agentic AI for critical operations.

"The biggest barrier to enterprise adoption of agentic AI isn't ROI - it's trust," said Siva Moduga, Co-Founder & CEO, Supervity AI. "Enterprises want AI systems that can autonomously execute critical operations, but they cannot compromise on security, data sovereignty, or performance. Supervity is building self-driving enterprise operations powered by AI Employees, and that requires a secure-by-design infrastructure foundation. Cloudbrink enables us to encrypt and isolate AI traffic end-to-end, protect access to private enterprise systems, and do so without introducing latency or operational friction."

Cloudbrink said the updated AI platform will be available next month.