Note Before Downloading, These Are AI Browser Shortages

AI ChatGPT Atlas browser display (photo: dock. OpenAI)

JAKARTA Currently, many Artificial Intelligence (AI) companies are launching AI-based browsers. The company is racing to beat the dominance of Google Search in the world.

Usually, this browser offers advanced features. It is even claimed to be more sophisticated than the usual browser. Although it looks interesting, there are some shortcomings you need to know before downloading the AI browser, according to Makeuseof.

AI-powered browsers often face serious security vulnerabilities called prompt injection attacks. This attack occurs when the website inserts hidden instructions followed by AI.

The core problem lies in the way AI agents process web pages. This agent reads all the text on the page as a whole, including text that is not visible to the user. AI can view the hidden text as part of the user's command.

This attack is a real problem because AI agents are designed to understand the entire page at once. If a site hides a line of texts, the AI browser can consider it a real order.

Thus, the site can order AI to ignore user requests or display unwanted information. Even OpenAI recognizes this vulnerability in the browser they just launched.

They stated that the Atlas ChatGPT was still vulnerable to prompt injection attacks. Despite its sophisticated concept, the Agent Mode feature in the AI browser is often less effective in practice.

Users report that the feature tends to moderate or summarize text randomly. The problem is, Agent Mode is often slow and even blocks instead of saving time.

However, the biggest issue is the security consequence of the vulnerability. The AI browser can be fooled by attacks such as phishing emails hidden on the page. This scam is a fatal failure that is very detrimental to users.