Grok Vulnerability Exposes User Chat Data via ‘Crypto Context Injection’

The significance here is less about Grok itself than about the class of attack: prompt-layer manipulation that turns an AI assistant into a data-exfiltration path. Crypto users are a natural target because chat histories with AI tools often carry wallet addresses, counterparties, and deal context that never belonged in a general-purpose model. Whether this remains a disclosed flaw or becomes a template for similar injections across other assistants is the open question worth watching.