How to fool AI Browser with the Hashjack "#" method

Near the end of the year, many readers will see that one of the breakthroughs of artificial intelligence or AI, besides the popular chat tool ChatGPT, another exciting news of the year is inevitable, an AI web browser with many developers competing to create, but soon a variety of hacking methods have been discovered.

According to a report by the website, Cyberpress has mentioned the discovery of a new type of AI web browser hacking method, called "HashJack," with the use of the "Hash" or "#" mark (the same one used to create the Hash Tag that many people are familiar with), and put it on a URL that points to a real website to hide the AI command (Prompt) behind the mark. What is hidden behind the mark is not sent to the server of the website, but runs only on the running machine (Client Side), resulting in a web browser scam that Going to enter a secure, reliable website, but running dangerous commands on the victim's machine at the same time. This method has the results of several AI web browsers, including Comet, Copilot tool on Edge, and Gemini on Chrome.

By way of a research team from Cato CTRL, a cybersecurity expert, the results of the test were discussed by that method, detecting six different scenarios in the attack by that method:

Callback Phishing will trick the victim into returning to a hacker-controlled number or a Whatsapp chat group created like an official group to deceive the victim in the next step.

Data Exfiltration Attacks are the use of this code to attack web browsers that use AI as an automatic substitute, or Agental AI. For example, Comet instructs web browsers to send financial information and sensitive information to hackers automatically.

Misinformation Attacks is the use of AI to enter credible-looking fake news through the victim's web browser.

The AI Malware Guidance is a step-by-step command for the AI to install malware on the victim's machine.

Medical Harm Attacks are the use of such commands to enable the AI to display fake medical information, such as changing the amount of medication required to harm the victim of misuse.

Credential Theft is an AI insert a fake login link that takes the victim to a fake website that is set to steal the password from the victim.

After the research team completed the experiment, it immediately contacted different AI developers, each with a different response.

It was announced by Microsoft on August 20, and the leak was successfully filled on October 27.

Google made the announcement on October 3, but so far it hasn't been announced if it's fixed or not.

Perplexity rejected the report at an early date, but changed its response during October 10 by declaring it a serious danger vulnerability, or Critical Severity (P1), and announced that the problem had been successfully resolved on November 18.

# Recap 2025 # Take care of yourself # Open budgetIT Matters Included# IT should know # freedomhack