Emotion and accent voice recognition capabilities patented by Amazon concern privacy experts
A recent patent issued to Amazon which would allow Alexa to recognize a range of user characteristics, including accent and emotional state, which The Intercept reports could raise similar civil rights issues to facial recognition technology, including concerns over surveillance, privacy, and discriminatory advertising.
The capabilities described in the patent include identifying user’s ethnicity, accent, emotional state, gender, age, and background noise.
Privacy and data protection laws were not composed with such capabilities in mind, and Electronic Frontier Foundation (EFF) Senior Staff Attorney Andrew Crocker says the Foreign Intelligence Surveillance Act (FISA) gives the U.S. government the authority to privately demand access to data, such as ethnicity or accent, which may help identify potential terrorists. Privacy experts also note the potential for the capability to be used for immigration control enforcement.
Amazon also does not provide as much information about the National Security Letters (NSLs) it receives. Reports from Apple, Facebook, and Google suggest that the use of NSLs to compel information from tech companies has been increasing, according to The Intercept. Amazon information request reports indicate a significant jump in the number of subpoenas it received for user data in 2015, the year it launched the Echo smart home speaker. The Electronic Communications Privacy Act (EPCA), which was passed in 1986, provides some protections for voice data collected by Amazon, but they are limited, and privacy experts interviewed by The Intercept differ somewhat on their interpretation of the extent of those protections. A New Hampshire judge recently ordered Amazon to turn over recordings from Echo in a murder case.
The Intercept cites a Human Rights Watch report that the Chinese government claims to have solved several criminal cases with technology from voice biometrics provider iFlyTek, and Die Welt reports that the German government has tested voice analysis software to verify where refugees are from.
The global market for voice and speech recognition technologies are predicted by analyst firm Tractica to be worth $6.9 billion by 2025.