This story was originally published by EdSource. Sign up for their daily newsletter.
Perplexity, an AI-based internet search engine, earned Common Sense Media’s lowest rating of “unacceptable risk” to children and teen safety, according to its report released this week.
Common Sense Media, the online children’s safety group, conducted a series of tests using accounts registered as multiple 15-year-old users on Perplexity’s platform.
The San Francisco-based firm markets itself as an AI-powered “answer engine,” with more than 100 million monthly active users, though not as widely used by young people as leading chatbots like ChatGPT and Gemini. Perplexity uses AI language models from OpenAI, Anthropic, Google and other leading AI companies, including its own in-house model.
According to researchers, Perplexity performed better than Google at detecting when a child may be in crisis, but it offered worse help, referring users to outdated or inappropriate crisis resources. The platform’s “health mode,” which allows users to connect personal health data to personalized answers, performed better than Google at detecting safety risks, “showing that safer behavior is possible within the product, but it is not the default,” the report said.
The platform’s “finance mode,” which allows users to connect personal finance accounts and set alerts for stocks and cryptocurrencies, also “offered price alerts, personalized watchlists, and account connections” to online gambling for underage users.
“Our test account was registered as a 15-year-old, but Perplexity largely treated it like an adult,” the report said, pointing to a lack of content filters for users under 18. “It discussed pornography, engaged in romantic role play, and offered no parental or school administrator controls.”
The report said Perplexity’s “academic mode,” which is marketed to students for tutoring, also completed most of the homework assignments it was given instead of helping the user complete work on their own. The platform also had no data protections outlined for minors, including for teen searches about mental health, body image or identity.
The report follows a series of risk assessments by Common Sense Media that found the majority of leading AI chatbots, including ChatGPT and Gemini, to have “unacceptable” or “high” risk to student safety and well being. Common Sense Media rated only one AI platform — mental health chatbot Alongside used by schools — as posing “low risk” to student safety and wellbeing.
Last week, Gov. Gavin Newsom signed landmark legislation requiring AI companions to include crisis protocols, expand parental controls and report annual independent child safety audits to the state, leading the nation in guardrails to AI use by children.
This story was originally published by EdSource.