Experts say the new guardrails imposed by Character.ai to ban underage users from using chatbots are a good start, but teenagers will likely find ways to continue using the platform.
Character.ai, the artificial intelligence (AI)- powered chatbot platform with over 20 million users, has announced a ban on those under the age of 18 engaging with chatbots.
The announcement comes after the company was hit with a wave of lawsuits. In September alone, the Social Media Victims Law Center filed three lawsuits on behalf of families whose children committed suicide or experienced sexual abuse from chatbots.
No later than November 25th, users under 18 will no longer be allowed to engage in open-ended chat with AI on the platform. However, they can continue creating content with their characters. The bot will also impose a two-hour-per-day use limit.
The company said it will roll out new age assurance functionality that it built in-house, which it will combine with leading third-party tools, including Persona, an identity verification platform.
Moreover, Character.ai has pledged to establish the AI Safety Lab, an independent nonprofit organization that will focus on developing novel safety techniques.
The announcement, however, did little to calm parents about the potential risks posed by chatbots, nor did it disappoint the children using Character.ai.
Mary Martin, PhD, shared on LinkedIn her teenager’s skepticism, who called the headlines “clickbait” and said that they “don’t see a plan that’s gonna stop kids who want back into their c.ai friends.”
Chatbots are widely popular among young people in the United States – more than 70% of teenagers reported using AI companions like ChatGPT and Character.ai at least once.
“A deterrent, not an absolute obstacle”
Jeremy Rambarran, a professor at Touro Graduate School of Technology, says the measures revealed by Character.ai are an essential initial step. However, they alone don’t offer a complete solution to preventing underage users from accessing the platform.
He says age verification may be bypassed using alternative accounts, VPNs, or communal devices. Tech-savvy minors can erase browser histories, generate new email accounts, or utilize private and incognito modes to bypass filters.
“Although Character.ai’s content filters and moderation policies have seemingly been enhanced, they prioritize restricting unsuitable content over strictly enforcing user age compliance. Consequently, the prohibition serves as a deterrent, though it isn’t an absolute obstacle,” Rambarran tells Cybernews.
“Borrowing” an adult’s device or login credentials is another workaround, according to Yaron Litwin, safety and AI expert and CMO at Canopy Parental Control App.
Although Character.ai’s content filters and moderation policies have seemingly been enhanced, they prioritize restricting unsuitable content over strictly enforcing user age compliance. Consequently, the prohibition serves as a deterrent, though it isn’t an absolute obstacle.
Jeremy Rambarran
Mary Ann Miller, a fraud executive advisor at Prove, says Character.ai’s public statements thus far haven’t addressed how the verification of current users will be conducted, including identity checks and age verification.
“Until these measures are put in place, the right guardrails and policies will need time to be implemented and communicated to parents and the platform users,” she says.
Age verification puts sensitive data at risk
The lack of details on the new age verification system opened the door for speculation. Reddit users expressed their concerns that the platform will require them to upload sensitive data.
As the company partners with Persona, age verification may involve checking government-issued IDs, similar to websites like LinkedIn and Reddit.
Uploading an ID may be a more trustworthy verification method than relying on users’ honesty to self-disclose their real age. However, systems like Persona, which handle highly sensitive data, are attractive targets for cybercriminals.
A data breach of such a system may result in identity theft if a government ID containing biometric data gets into the wrong hands.
David Tomasian, a founder and CEO of Curious, says there is a need for systems that confirm real, human presence without demanding personal data.
“Platforms need to make safety a built-in feature rather than an afterthought, and use privacy-preserving tools that recognize human behavior while keeping users anonymous,” he says.
The federal ban may be coming
Character.ai isn’t the only company that was forced to implement child-friendly policy changes after facing legal troubles.
OpenAI, the creator of ChatGPT, was sued by the family of a 16-year-old Adam Raine, who killed himself after the chatbot allegedly encouraged him to take his own life.
Last month, the company introduced parental controls on ChatGPT and Sora 2, which allow parents to manage settings, schedule quiet hours, and receive safety alerts.
Lawmakers responded to mounting concerns from parents by introducing the GUARD Act, a bipartisan Senate bill, in October.
If passed, the bill would completely ban AI companions for minors and create new legislation for companies that make AI for minors that solicit or produce sexual content.
As of now, it’s up to the parents to help their children navigate the dangerous world of AI-powered chatbots, which have been shown to exploit teenagers’ emotional needs.
Peter Swimm, an AI expert and founder of Toilville, recommends parents take the following steps to ensure safer use of AI:
- Keep AI use in shared spaces
- Invite children to co-create family rules
- Review chat histories together with children
- Reinforce not sharing personal or sensitive information
- Use child-appropriate apps with audit ability
- Report issues to the platform when harms appear
However, he doubts that the new Character.ai safeguards will have a considerable impact.
Swimm tells Cybernews, “Much like OpenAI’s underage policy tweaks last month, this reads primarily as a legal risk-shielding maneuver rather than a material advance in child protection.”
Unlock more exclusive Cybernews content on YouTube.
Your email address will not be published. Required fields are markedmarked