Drawing on new research into emerging AI-powered browsers, Cyberhaven Labs highlights how unresolved security challenges, such as prompt injection, could enable data leakage and manipulation inside corporate environments.
MOUNTAIN VIEW, Calif., Oct. 30, 2025 /PRNewswire/ — Cyberhaven, the leader in AI-powered data security, today released new research around the corporate adoption of OpenAI’s new ChatGPT Atlas, an AI-powered browser that allows users to interact with ChatGPT on any browser tab directly. The comprehensive analysis from Cyberhaven Labs draws on actual usage patterns covering millions of users. The findings reveal rapid enterprise adoption and underscore emerging risks associated with a new class of “agentic browsers,” also known as “browser agents.”
Since the release of ChatGPT Atlas on October 21, Cyberhaven Labs tracked its adoption in corporate environments and investigated associated security vulnerabilities, detailed in their latest research.
Early data shows that Atlas is spreading through enterprises at an unprecedented rate, outpacing prior AI launches in both speed and scale. The analysis explores adoption trends by industry, workforce penetration, and comparisons with competing agentic browsers.
Adoption Statistics
- 27.7% of enterprises have at least one employee who downloaded ChatGPT Atlas, with some organizations seeing up to 10% of their workforce using it.
- 1.7% of corporate macOS endpoints now have ChatGPT Atlas installed – equivalent to 17 out of every 1,000 devices.
- By industry, 67% of users were in technology, 50% in pharmaceuticals, and 40% in finance.
- Compared with Perplexity’s Comet browser (released July 9, 2025), ChatGPT Atlas has seen 62 times more corporate downloads.
- During the week of Atlas’s release, Comet downloads surged sixfold, signaling renewed interest in agentic browsers overall.
According to Cyberhaven’s broader 2025 AI Adoption and Risk Report, ChatGPT remains the most widely used AI tool in the workplace, a key driver of Atlas’s rapid enterprise penetration.
Agentic browsers like ChatGPT Atlas introduce powerful new capabilities but also significant data loss prevention challenges for corporate data.. A leading concern is prompt injection, a still-unsolved security problem that could allow malicious web content to manipulate or exfiltrate sensitive data.
Cyberhaven’s endpoint agent and browser extension can track data lineage within ChatGPT Atlas, helping organizations protect sensitive data from both malicious insiders and compromised AI agents.
To learn more about agentic browser risk and AI security, join the Cyberhaven Unlock Series Webinar: Security for AI on November 25. Register here.
About Cyberhaven
Cyberhaven is reimagining data security. Until now, data security products have been limited to scanning data content or looking for specific user actions. Our AI-enabled data lineage technology analyzes billions of workflows to understand every piece of data within an organization, identify when it’s at risk, and take action to protect it. To learn more, visit cyberhaven.com.
SOURCE Cyberhaven