‘Sexualized’ AI Chatbots Pose Threat to Kids, Warn Attorneys General in Letter

The Nationwide Affiliation of Attornneys Normal (NAG) has written to 13 AI corporations, together with OpenAI, Anthropic, Apple and Meta, which requires stronger ensures to guard youngsters towards inappropriate and dangerous content material.

It warned that youngsters have been uncovered to sexually suggestive materials through “flirty” AI chatbots.

“Exposing youngsters to sexualized content material is indefensible,” wrote the generals of the attorneys. “And conduct that will be unlawful – and even prison – whether it is achieved by folks, is just not apologetic, just because it’s achieved by a machine.”

The letter additionally made comparisons with the rise of social media, and mentioned that authorities businesses didn’t do sufficient to emphasise the methods wherein the youngsters negatively influenced.

“Social media platforms have prompted lots of harm to youngsters, partly as a result of authorities wakes canine haven’t achieved their work rapidly sufficient. Teached classes. The potential harm from AI, such because the potential advantages, dwarf the influence of social media,” the group wrote.

Using AI in youngsters is widespread. Within the US, a examine of widespread sense media found Zeelen-Windenvermark that seven out of ten youngsters had tried generative AI from 2024. In July 2025 it turned out that greater than three-quarters of AI companions used and that half of the respondents mentioned they usually trusted it.

Different nations have seen comparable developments. In the UK, a survey final yr by Regulator Ofcom confirmed that half of the web 8-15 yr olds had used a generative AI software within the earlier yr.

The rising use of those instruments has sparked mounting concern from mother and father, colleges and youngsters’s rights grours, who level to danger ranging starting from sexual suggestive “flirty” chatbots, ai-generated little one, groom, bleyining andbullying andbullying and beadying andbullying andbullying and beadying andbullying andbullying and-based andbullyining and Breullying and Breullying, and Poorly Understood Psychological Well being Impacts.

Meta not too long ago got here beneath particular hearth after leaked inner paperwork revealed that his AI assistants have been allowed to “flirt and play romantic function play with youngsters”, together with these as younger as eight. The recordsdata additionally confirmed coverage that chatbots allowed youngsters to inform their “youthful type is a murals” and describe them as a “darling”. Meta later mentioned it had eliminated these tips.

Naag mentioned that the revelations for Advocaten-Normal “revolted” due to this obvious contempt for the emotional well-being of youngsters “and warned that dangers weren’t restricted to meta.

The group talked about lawsuits towards Google and character.ai who claimed that sexualized chatbots had contributed to the suicide of an adolescent and inspired one other to kill his mother and father.

Among the many 44 signatories, Tennessee was the legal professional -general Jonathan Skrmetti, who mentioned that corporations can’t defend the coverage that sexualized interactions with minors normalize.

“It’s one factor for an algorithm to wander – that may be solved – however it’s one other for individuals who run an organization to just accept tips that give in -affirmative care permission,” he mentioned. “If we can’t maintain innovation away from harm youngsters, that isn’t progress – it is a plague.”

Decrypt has contacted however not but heard of all AI corporations talked about within the letter.