OpenAI has shipped a security update to ChatGPT Atlas aimed at prompt injection in AI browsers, attacks that hide malicious instructions inside everyday content an agent might read while it works.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results