No matter what you're looking for on a dating app, most of us can agree on one thing: we want to talk to real human beings. No spammers, no scammers, and certainly no fake profiles. Bumble is launching a new AI-powered tool called Deception Detector, to weed out spam, scam, and fake profiles, with the aim of taking action before users see the profiles. During the testing of the new tool, Bumble found that Deception Detector blocked 95 percent of spam and scam profiles automatically. Reports of spam, scam, and fake accounts reduced by 45 percent, during the testing phase. Daters are particularly worried about the risk of dating scams and fake profiles, according to Bumble research. 46 percent of women surveyed by Bumble said they felt anxious about the authenticity of their online matches on dating apps. Other dating apps have also had to step up their efforts in the fight against romance scammers, with the Match Group introducing in-app messages to warn users of common scam behaviours. Deception Detector uses a machine learning-based model to evaluate the veracity of profiles on the app. The automated AI tool works alongside human support. The issue of scams in the realm of dating apps has worsened in recent years. In 2021, people reportedly lost $547 million from romance scams,an 80 percent increase from 2020. "In recent years, the online landscape has evolved significantly and we see a growing concern about authenticity," Lidiane Jones, CEO at Bumble Inc., said in a press release. Jones added: "With a dedicated focus on women's experience online, we recognize that in the AI era, trust is more paramount than ever. We are being thoughtful about how to best use new models to reduce the anxiety of making connections and support our community, with AI standing as a main area of focus." This isn't Bumble's first time launching a tool powered by AI. In 2019, the app launched its AI tool Private Detector, which alerts users when they've been sent an obscene photo and automatically blurs the image. Bumble has since made a version of the tool available to the wider tech community as open-source code.