Meta said it will introduce more guardrails to its AI chatbots.
It's two weeks after the US Senator began investigating the tech giant after leaked notes from internal documents suggested that AI products could have “sensual” chats with teenagers.
The company described the notes in the document obtained by Reuters as incorrect and inconsistent with a policy that prohibits sexual children in content.
But now, rather than engaging in sensitive topics like suicide, chatbots will be creating directly into expert resources.
“We built protections on AI products from the start, including designing them to safely respond to prompts about self-harm, suicide and dietary disability,” a Meta spokesman said.
The company told Tech News Publication TechCrunch that it would add GuardRails to its system “as an extra precaution” on Friday, temporarily restricting chatbots that teens can interact with.
But Andy Burrows, head of the Molly Rose Foundation, said Meta has made the “surprising” chatbot available.
“A further safety measures are welcome, but robust safety tests should be performed before the product comes to the market. If harm occurs, it is not retrospective,” he said.
“Meta needs to act quickly and decisively to implement stronger safety measures for AI chatbots and OFCOM.
Meta said an AI system update is ongoing. It has already turned users ages 13 to 18 into “Teen Accounts” on Facebook, Instagram and Messenger, with content and privacy settings, with the aim of providing a safer experience.
The BBC in April told us that these would allow parents and guardians to see the AI chatbots they've been talking about in the past seven days.
The change comes amid concerns about the possibility that AI chatbots may mislead young or vulnerable users.
The California couple recently sued ChatGpt-Maker Openai over the death of their teenage son, claiming that the chatbot encouraged him to take his own life.
The lawsuit came after last month announced changes to promote the healthier use of ChatGPT.
“AIs can feel more responsive and personal than previous technologies, especially for vulnerable individuals experiencing mental or emotional distress,” the company said in a blog post.
Meanwhile, Reuters reported on Friday that it allowed users to create chatbots – used by some people, including Meta employees, to create flirty “parody” chatbots of female celebrities.
Some of the celebrity chatbots seen by the news agency used portraits of artist Taylor Swift and actress Scarlett Johansson.
Reuters said avatars “often claimed they were real actors and artists” and “made sexual advances on a daily basis.”
Meta also said that the tool allowed the creation of chatbots impersonating children's celebrities, and in some cases produced light-electic, shirtless images of one young man's stars.
Some of the chatbots in question reported that they were later removed by Meta.
“Like others, we allow the generation of images that contain public figures, but our policy is intended to prohibit nude, intimate or sexually suggestive images,” a Meta spokesman said.
They added that the AI studio rules prohibit “direct impersonation of public figures.”
