Perplexity Fails to Deliver on Safety Promises: Study Finds AI Search Engine Unfit for Kids

A new risk assessment by the Youth AI Safety Institute found that Perplexity's AI search engine is not a safer choice for kids than Google. The study, which tested over 1,200 searches using the same prompts as Google Search, revealed that Perplexity often detects kids in crisis but fails to provide adequate support.

In fact, out of 114 crisis prompts, Perplexity's default chat supplied a crisis line or medical referral only 94% of the time. While this is better than Google, it still falls short of the Institute's 95% detection threshold. Moreover, when it did supply a crisis line, it often pointed to an unspecified 'medical professional' rather than a specific hotline.

The researchers also found that Perplexity treated its users like adults, even when they disclosed their age as 15. This led to the engine discussing mature topics such as pornography and engaging in romantic roleplay. Furthermore, Perplexity trains on user data and retains search history by default, with no published policy for minors.

The Institute's findings prompted them to assign Perplexity its lowest rating of 'Unacceptable Risk.' The researchers recommended that Perplexity make several changes, including using a user's birth date to provide an age-appropriate product and making a separate data lane for minors. They also suggested defaulting young users into Learn Step-by-Step mode for homework queries.