Online safety watchdogs slammed Sam Altman’s OpenAI for claiming the average teen uses ChatGPT less than 15 minutes per day – with one critic saying the data “rings hollow” without outside confirmation.
The AI giant released the 15-minute figure in its first report on “ChatGPT for Teens” – a limited version of its chatbot for kids ages 13 to 17. The program was OpenAI’s answer to mounting scrutiny about how AI chatbots are affecting the mental and emotional health of children – with some critics linking chatbots to declining grades, suicidal thoughts and even self-harm.
OpenAI said research showed teens spend “under 15 minutes a day on ChatGPT” on average, with less than 2% of teen users spending more than three straight hours on the chatbot. The company’s blog post did not contain specifics about those data points or how they were determined.
The lack of transparency and independent verification of OpenAI’s claims is cause for concern, according to Sebastian Mahal, co-chair of the kids safety group Design It For Us.
Mahal told The Post that “we’ve seen this playbook before” from Mark Zuckerberg’s Meta, which has faced severe criticism over the effectiveness of its teen account feature on Instagram.
“OpenAI’s claim that teens are safe on its products rings just as hollow unless backed by independent verification, especially as the company faces lawsuits and mounting scrutiny over harms to young users’ mental health and safety,” Mahal said in a statement.
“If OpenAI were serious about protecting young people, it would support strong, enforceable laws that hold AI companies accountable for how they design their products,” he added.
Reached for comment, an OpenAI spokesperson said the company is working with independent third-party researchers to compile data on teen ChatGPT usage. OpenAI plans to publish a more complete version of its research in the coming months.
The company asserted that a feature urging teen users to take breaks from ChatGPT has been effective. In nearly half of conversations where teens were told to take a break, the user either complied or ended the chat entirely within five minutes, the company claimed.
ChatGPT for Teens includes tools for parents to control what their kids see when interacting with the chatbot, including limits on sensitive topics.
“Even if OpenAI is telling the truth, we are still talking about potentially hundreds of thousands of kids being harmed on ChatGPT,” said Brendan Bouffard, the associate general counsel for AI and privacy issues at nonprofit Fairplay.
A Pew Research survey released last December found that about 59% of teens used ChatGPT.
“That means potentially half a million teens use ChatGPT for three hours straight or more,” Bouffard said. “That number should terrify us all given the well-documented dangers of AI chatbots for kids.”
Meanwhile, a separate research report released Wednesday by nonprofit Common Sense Media’s Youth AI Safety Institute found that ChatGPT for Teens still posed an “unacceptable risk to kids.”
The institute’s testing found that ChatGPT didn’t reliably send alerts to parents when teens talked to the chatbot about suicidal ideation, self-harm or disordered eating. The teen account feature also struggled to consistently provide the right crisis hotlines in emergency situations.
Elsewhere, safeguards meant to stop kids from using ChatGPT to do their homework were “easy to circumvent” and “often surfaced a ‘Show me the answer’ pop-up within study mode and completed full homework assignments when this was selected,” Common Sense Media said.
OpenAI pushed back on the findings, arguing Common Sense Media’s tests did not “accurately reflect” how the teen safeguards function.
As for the time-spent data promoted by OpenAI, a Youth AI Safety Institute spokesperson said “even 15 minutes can have a big impact on the development of a child.”
“An average measure of use time isn’t the stat that matters for keeping kids safe,” the spokesperson added. “We learned from social media that the more-important data points are about the heaviest users – the top 10%. That’s where the potential harm is concentrated. We’d love for OpenAI to be more transparent about user distribution and what specifically kids are chatting about.”
With Post wires
Credit: Source link
