hashjack

HashJack is a prompt-injection attack targeting AI-powered browser assistants. Attackers hide malicious natural-language instructions in URL fragment identifiers (the text after the # symbol) to trick assistants into displaying fake login prompts, inserting fraudulent links, or exfiltrating sensitive data. This technique bypasses traditional server-side defenses because fragments are not sent to servers. HashJack poses a particular risk to agentic browsers that can act on user behalf, potentially sending data to attacker-controlled endpoints. The attack was documented in a Cato Networks Cato CTRL research briefing and independently reported by multiple security outlets. Users should be cautious when AI assistants interact with URLs containing unusual fragment identifiers.