“Alexa, steal my passwords.” It’s not a phrase a user is likely to utter, but security researchers in Germany have shown that it’s possible for malicious apps — Alexa “skills” and Google Home “actions” — to launch phishing attacks on users, forward the compromised credentials to criminals, and do it all in apps approved for use by the voice-assistant giants.

Security Research Labs, a white-hat research organization, developed a total of eight apps, four each for Amazon Alexa and Google Home, that masqueraded as horoscope checkers or a random number generator. The apps triggered malicious actions based on action words like “stop,” while continuing to operate after users thought they had closed.