A new vulnerability, 'Whisper Leak,' could allow attackers to infer topics discussed with AI chatbots by analyzing network traffic patterns, Microsoft has warned. The company also highlighted the risks of surveillance by oppressive governments targeting sensitive discussions