Perplexity's Comet and OpenAI's ChatGPT Atlas are at the forefront of agentic AI browsers. Brave's research highlights prompt injection risks in agentic AI browsers that could allow unauthorized actions by malicious websites, prompting concerns over user security.