随着人工智能可穿戴录音设备日益普及，隐私保护与监听技术之间的对抗也在不断升级。初创公司Deveillance在3月宣布开发反录音设备Spectre I，旨在防止他人在用户不知情的情况下窃听[1]。公司创始人Aida Baradari表示，她对AI可穿戴录音设备激增感到担忧[1]。据传苹果正在开发AI pin或pendant，将其作为iPhone的常时眼睛和耳朵[1]。
学术界和科技企业多年来在语音处理领域持续开展研究。俄亥俄州立大学的DeLiang Wang使用神经网络训练了数百小时的人类语音，教会模型识别语音模式[1]。微软自2020年以来每年举办Deep Noise Suppression Challenge[1]。与此同时，2023年南京大学的Ming Gao团队开发了MicFrozen，一种实时生成超声波"反语音"的反制设备[1]。这种反复的技术迭代表明，监听方和防护方正在进行持续的对抗。
专家指出，AI可穿戴设备有可能绕过直接录音，转而通过唇读或水表面振动等其他方式恢复语音[1]。
Startup Deveillance announced the development of Spectre I in March, a device designed to prevent unauthorized audio recording.[1] The initiative was led by founder Aida Baradari, a recent graduate concerned about the proliferation of AI-powered wearable recording devices.[1] Apple is reportedly developing an AI pin or pendant that would serve as a constant audio and visual interface for the iPhone.[1]
The emergence of such surveillance-capable devices has triggered an escalating technological competition. Researchers at Ohio State University, led by DeLiang Wang, have trained neural networks on hundreds of hours of human speech to enable AI systems to identify and recover vocal patterns.[1] Microsoft has been holding annual Deep Noise Suppression Challenges since 2020 to advance voice recovery capabilities.[1] In response, a team from Nanjing University led by Ming Gao developed MicFrozen in 2023, a countermeasure device that generates real-time ultrasonic "anti-speech" signals.[1]
The technological landscape has expanded beyond traditional audio blocking, as AI wearable devices may potentially circumvent recording through alternative recovery methods such as lip-reading or detecting vibrations on water surfaces.[1] The scope of surveillance capabilities has extended to law enforcement, as the FBI secretly operated an encrypted phone company called Anom from 2018, selling 12,000 devices for monitoring purposes.[1]