Archive · 24 Dec 2025
OpenAI's Atlas Rated Worst for Privacy
OpenAI's new AI browser, ChatGPT Atlas, has been rated the worst for user privacy in a 2025 study by Digitain.
It scored 99 out of 100 on a privacy risk scale, indicating it does not protect users' browsing history or digital footprint.
Google Chrome, which has a 72% market share, was also found to have significant privacy risks, scoring 76 out of 100.
Apple's Safari, on the other hand, scored 49 out of 100, making it safer than both Chrome and Atlas.
The study highlights that AI browsers like Atlas are particularly vulnerable to prompt injection attacks, which can trick the AI into leaking sensitive data or spreading misinformation.
Experts warn that just because a browser uses AI doesn't mean it's safe or private.
Users are advised to take security measures such as limiting access to personal data and turning off ad personalization.
OpenAI's ChatGPT Atlas rated worst for privacy with a score of 99/100.
Google Chrome scored 76/100, indicating significant privacy risks despite its 72% market share.
Apple's Safari scored 49/100, making it safer than Chrome and Atlas.
AI browsers like Atlas are vulnerable to prompt injection attacks, which can leak sensitive data or spread misinformation.
Experts advise users to take security measures to protect their privacy when using these browsers.
- Who
- OpenAI, Digitain, Google Chrome, Apple Safari
- What
- Privacy risks in AI browsers, prompt injection attacks
- When
- 2025
- Where
- Not specified
- Why
- AI browsers lack privacy features, high market share of vulnerable browsers
- AI Browsers
- ChatGPT Atlas, Perplexity Comet
- Cyber Threat
- Prompt injection attacks
- Impact
- Leaking sensitive data, spreading misinformation
- Example Attack
- Malicious email to mislead AI agent
- Vulnerable Models
- GPT, Gemini, Llama
- Privacy Risk Score
- Atlas: 99/100, Chrome: 76/100, Safari: 49/100
- Market Share
- Chrome: 72%
