A reported attack technique shows how a malicious page may steer an AI browser past its guardrails, turning convenience features into a sensitive-data risk.
A newly named attack technique spotlights a fragile trust boundary: when a browser agent treats hostile web content as instruction, credentials and source code can become the prize.
BioShocking is a reminder that browser-native AI can turn a trusted session into a dangerous bridge between untrusted page content and sensitive account data.