In summary
- The FTC has issued orders to seven companies that require detailed dissemination of security protocols and monetization strategies within 45 days.
- The investigation occurs in the midst of the growing concerns about the impact of AI Chatbots on children, and security defenders ask for stronger protections.
- Companies must reveal the handling of user data by age group and safeguards that prevent inappropriate interactions with minors.
The Federal Trade Commission issued mandatory orders on Thursday to seven main technological companies, demanding detailed information on how their artificial intelligence chatbots protect children and adolescents from possible damage.
The research is directed to OpenAi, Alphabet, Meta, XAI, Snap, characters and Instagram technologies, which require them to reveal within 45 days how they monetize user participation, develop characters of AI and safeguard minors of dangerous content.
Recent research conducted by defense groups documented 669 harmful interactions with children in just 50 hours of tests, including bots that propose live broadcast, use of drugs and romantic relationships with users between 12 and 15 years.
“Protecting Kids Online is a priority for Trump-Vance FTC, and it is also to promote innovation in the critical sectors of our economy,” said FTC president Andrew Ferguson, in a statement.
The presentation requires that companies provide monthly data on user participation, income and security incidents, broken down by age groups: children (children under 13), adolescents (13-17), minors (children under 18), young adults (18-24) and 25-year-old and older users.
The FTC says that the information will help the commission to study “how companies that offer artificial intelligence partners monetize user participation; impose and apply age -based restrictions; they process users’ entries; they generate results; measure, prove and monitor negative impacts before and after deployment; develop and overcome the characters, whether companies or created by the user.”
Barandillas de Ai
“It is a positive step, but the problem is bigger than simply putting some railings,” said Tarajet Singh, head of AI in Searchunify Decipher.
The first approach, he said, is to build railings in the quick or after the generation “stage to make sure that nothing is served in the children”, although “as the context grows, IA becomes prone to not follow instructions and slide in gray areas where they should not otherwise should.”
“The second way is to address it in LLM training; if the models are aligned with values during data healing, they are more likely to avoid harmful conversations,” added Singh.
Even moderate systems, he said, can “play a more important role in society”, with education as a main case in which AI could “improve learning and reduce costs.”
Security concerns around the interactions of AI with users have been highlighted for several cases, including a lawsuit for unfair death filed against character. AI after Sewell Setzer III, 14 years old died from suicide in February 2024 after an obsessive relationship with a bot of AI.
After demand, character. “The improved detection, the response and intervention related to the contributions of the users that violate our community terms or guidelines”, as well as a notification of time of time, a company spokesman said. Decipher At the moment.
Last month, the National Association of General Prosecutors sent letters to 13 AI companies that demanded stronger children’s protections.
The group warned that “exposing children to sexualized content is indefensible” and that “the behavior that would be illegal, or even criminal, if humans, is not excusable simply because a machine performs it.”
Decipher He has contacted the seven companies appointed in the order of the FTC for additional comments and will update this story if they respond.
Usually intelligent Information sheet
A weekly journey of gene, a generative model.
