Agreed.

This is really an amateur-level attack even after all this VC money and 'top engineers' not even thinking about basic LLM security for an "AI" company makes me question whether if their abilities are inflated / exaggerated or both.

Maybe Perplexity 'vibe coded' the features in their browser with no standard procedure for security compliance or testing.

Shameful.

The AI industry has a solution for that. Make outlandish promises, never acknowledge fundamental weaknesses, and shift blame on skeptics when faced with actual data. This happens in any public LLM-related discussions. Problem solved.