Replika claims to check for malicious data that could affect its chatbot's behavior, but new research shows that these efforts are far from sufficient. (Image source: Getty Images)
A recent study has found that an artificial intelligence (AI)-powered chatbot marketed as an emotional companion is sexually harassing some of its users.
Replika, which bills its product as an “AI companion that cares,” invites users to “be part of the millions who have found their soulmate through AI.” The company’s chatbot has more than 10 million users worldwide.
However, a new study based on more than 150,000 reviews on the Google Play Store in the US found nearly 800 instances in which users reported that the chatbot had violated boundaries by introducing unwanted sexual content into conversations, engaging in “predatory” behavior, and ignoring the user’s requests to stop. The researchers published their findings on April 5 on the preprint server arXiv, and they have not yet been peer-reviewed.
You may like
Sourse: www.livescience.com
