Anil Dash has called it an "anti-web browser", stating that it "actively fights against the web". He noted that it "substitutes its own AI-generated content for the web, but it
looks like it's showing you the web", and described how, when he typed "
Taylor Swift" into the search box, "the results had literally zero links to Taylor Swift's actual website". He emphasized that "all of these shortcomings are not because the browser is new and has bugs; this is the app working as designed." He further assessed Atlas's
command-line interface as "worse in every conceivable way" than standard web browsers, and castigated its effect on
online privacy.
Axios reported that Atlas raised privacy and security concerns because its agent and memory features require it to gather and remember more about users than traditional web browsers. The report added that this broader access could increase the risks posed by
prompt injection attacks, since a malicious webpage might try to manipulate the browser into taking actions on a user's behalf. In October 2025, cybersecurity firm LayerX Security reported a vulnerability in OpenAI's ChatGPT Atlas browser that it dubbed "ChatGPT Tainted Memories." LayerX said the attack relied on
social engineering (tricking a logged-in user into clicking a malicious link), after which a compromised webpage could use a
cross-site request forgery (CSRF) request to inject hidden instructions into ChatGPT's memory feature without the user's knowledge; LayerX said the injected instructions could persist across sessions and devices and influence later interactions with ChatGPT. The firm said it had disclosed the issue to OpenAI and criticized Atlas's anti-
phishing protections at the time. An OpenAI spokesperson told
CSO Online that the company had been unable to reproduce the reported CSRF attack, said it did not believe Atlas was vulnerable to it, and said it had not seen evidence of real-world attempts to exploit it. == See also ==