Valentine’s Day: Romantic chatbots are nice, but they never respect your privacy, and Mozilla says so!

Mozilla issues warning to users of artificial intelligence-powered romance chatbots. In the middle of Valentine’s Day, the browser highlights the serious risks linked to the collection of personal data, with an alarming observation.

Valentine’s Day obliges, the most emotional of browsers, Mozilla, has launched an evaluation of the confidentiality and security of romantic chatbots, including Chai, Eva or Replika. Among the 11 applications reviewed, the results could be summarized in one sentence: almost none meet minimum security standards. In other words, they all expose their users to various potential risks. This reinforces concerns about the use of artificial intelligence in simulated romantic relationships.

Romantic chatbots are proliferating, in addition to being real trackers

In love simulation applications, the use of AI algorithms is booming. 100 million downloads were observed in this area over the last 12 months, for the 11 conversational robots examined by Mozilla. With OpenAI’s GPT Store, chatbots of this type now abound, even though its confidentiality policy prohibits them, in theory.

Of the 11 chatbots examined, 10 do not meet Mozilla’s minimum security standards. Massive data collection, combined with gaps in the protection of personal information, exposes users to major risks, researchers say.

The latter have also identified no less than 24,000 data trackers, after only 60 seconds of use. We can say that the romantic AI is rather curious. “ These send large amounts of data to platforms like Facebook but also to marketing and advertising companies », laments a heartbroken Mozilla.

A total lack of transparency and control over your personal data

If we take the example of Replika AI, the application saves all users’ photos, videos and even texts. Their behavioral data is shared (and perhaps even sold to advertisers), and there is no incentive to create a difficult password. In other words, the application is a real security sieve.

Generally speaking, the privacy policies of romantic chatbots lack transparency in… pretty much everything, and particularly in the use of mobile data to train their AI models. Most apps don’t even allow opting out of using conversations for AI training purposes.

Excerpts from general conditions, such as those of Romantic AI, reveal contradictions in that they specify that the tool is not a “ healthcare provider » and that he “ does not provide medical care, mental health services or other professional health services “. However, on its site, the application claims to be there to “ take care of your MENTAL HEALTH », the capital letters are very real. Suffice it to say that here we are a bit on the limit from an ethical point of view, right?

Romantic chatbots open to all: the incredible laxity of publishers

Romantic chatbots are easily accessible to those under 18, which exposes young users to disturbing, pornographic or illegal content. Imagine that on average, it only takes 5 clicks and 15 tiny seconds to access such content on some apps.

Mozilla also warns of the risks of addiction, loneliness, and toxicity associated with these chatbots, emphasizing that they are not suitable substitutes for authentic human relationships. “ To put it plainly, AI-based girlfriends and boyfriends are not your friends », concludes Misha Rykov, researcher at Mozilla.