AI-Powered Browsers: Navigating the Next Frontier of Enterprise Security

The browser wars are heating up — and this time it’s not about tab speeds or rendering engines, it’s about integrated AI. Microsoft, Perplexity and OpenAI are all racing to build browsers that proactively help you search, compare and buy, while Google is expanding its own AI-only search features.

What’s new?

Microsoft’s Edge gets Copilot Mode.

In a July 2025 blog post Microsoft announced that its experimental Copilot Mode transforms Edge from a passive window into a smart assistant. When enabled, the browser shows a single input box that combines chat, search and navigation and it "anticipates what you might want to do next" and "works with you as a collaborator". Copilot Mode can see the full context of your open tabs to help you compare options and make decisions.

Perplexity launches Comet.

TechCrunch reports that Perplexity’s Comet browser, released on 9 July 2025, puts the startup’s AI search engine front and centre. Its Comet Assistant can summarise emails and calendar events, manage tabs and navigate pages on your behalf. Comet is initially available to subscribers of the Perplexity Max plan】.

OpenAI is reportedly building a browser.

According to an exclusive Reuters report, OpenAI is preparing an AI-powered browser that includes a chat interface and AI agent integrations and aims to fundamentally change how consumers browse. Sources said the browser will keep some user interactions within a ChatGPT-like interface to capture more user data. With 500 million weekly ChatGPT users and Chrome’s three billion-plus users as a target, the move could put pressure on Google’s ad-driven Chrome dominance.

Google experiments with AI Mode.

Google’s AI Mode experiment uses a custom version of its Gemini 2.0 model to provide advanced reasoning and multimodal responses. It can answer complex, multi-part questions with follow-ups and uses a query fan-out technique that runs multiple related searches concurrently and merges the results. Google notes that AI Mode combines advanced model capabilities with the company’s best-in-class information systems.

Why CISOs, CIOs and IT leaders should care

These AI browsers are not just gadgets — they fundamentally alter how data flows through the browser. By design, AI agents need context. Microsoft’s Copilot Mode can "see all your open tabs" to provide multi-tab context. Perplexity’s Comet Assistant lives alongside your emails, calendar and web pages. Reuters notes that OpenAI’s browser is part of a broader strategy to capture data on users’ web behaviour. AI Mode in Google Search issues multiple related searches and aggregates results.

For security leaders, this means:

Expanded attack surface.

AI browsers are essentially agentic platforms with privileged access to browsing sessions, calendars and emails. Misconfigurations or malicious prompts could expose sensitive corporate information.

Data sovereignty concerns.

Reuters reports that OpenAI’s browser aims to use AI to change how consumers browse and to gain direct access to user data. CISOs must scrutinise how vendors collect and store browsing data and whether it can be used to train models.

Trust but verify.

While Microsoft promises that Copilot Mode is built to "the highest standards of security, privacy and performance", and Google says AI Mode is rooted in its core quality and ranking systems, enterprises still need to validate these assurances through contractual controls and technical assessments.

User education.

AI browsers may encourage users to offload tasks to agents (e.g., Comet can book flights or manage tabs). Employees must understand the risk of sharing sensitive information with assistants and avoid pasting confidential data into chat interfaces.

Navigating the AI browser era

1. Assess vendor AI implementations. Examine how each browser uses and retains data. Is AI processing done locally or in the cloud? Are there opt-in controls to disable agent access to other tabs or emails?

2. Update browsing policies. Treat AI browsers as new endpoints. Enforce least-privilege access; restrict which users can enable agentic features; and monitor traffic for signs of data exfiltration.

3. Test for prompt injection and data leakage. Work with security teams to simulate attacks against AI browsers, including malicious webpages that try to coerce the agent into leaking internal secrets or performing unintended actions.

4. Invest in observability. Ensure you have telemetry to see what prompts, actions and data flows occur through these new assistants. Without visibility, you can’t measure or mitigate risk.

Looking ahead

The rapid rise of AI-powered browsers signals a shift from passive browsing to active assistance. For CISOs, CIOs and heads of IT, the opportunity is to harness these innovations safely. By proactively evaluating AI browsers, updating governance and training users, organisations can benefit from productivity gains while preserving trust and compliance.

If your organisation is considering AI-powered browsing or building your own agentic tooling, let’s connect. I’m exploring ways to align AI innovation with robust security and would love to hear your perspective.