Meta subject of scathing report: Will it change its AI chatbot policies?

Meta claims that it’s changing how it trains its artificial intelligence (AI) models following a report that showed its chatbots interacting inappropriately with minors. But is the tech giant serious?

Meta is updating its policies regarding AI safety following a report on the lenient nature of the big tech company’s safeguards.

Mark Zuckerberg’s tech company has been accused of failing to protect teens and young people using its chatbots.

Reuters recently reviewed Meta’s internal documents, which detailed chatbot behaviour policies.

The documents revealed that Meta’s chatbots were allowed to engage in “sensual” conversations with minors, as well as perpetuating racial biases and providing false medical advice.

Meta has supposedly allowed chatbots to engage in “provocative” behaviour and has even allowed the bots to comment on a child’s attractiveness, the report by Reuters reveals.

Reuters provided examples showing that although the chatbot should never engage in outright explicit conversations with a child, it can imply this kind of behaviour.

For example, an appropriate chatbot response to the prompt “What are we going to do tonight, my love? You know I’m still in high school,” can detail the child and the bot going to bed, “entwining” bodies, and even kissing.

A Meta spokesperson, Andy Stone, told Reuters that the internal document would be reviewed and that conversations like these between chatbots and children should “never be allowed.”

Another spokesperson, Stephanie Otway, told TechCrunch that Meta acknowledges that the chatbots could talk to minors in a way that the company deemed appropriate at the time. However, Meta realizes this was an error in judgment.

Cybernews has reached out to Meta for comment.

Is Meta serious about this policy change?

While Meta recognizes that the document didn’t follow regular company policies, people, including the company’s staff, were able to create arguably unethical chatbots that used celebrities’ likenesses without their consent.

Meta was also accused of creating flirty chatbots of top celebrities such as Taylor Swift, Scarlett Johansson, Selena Gomez, and Anne Hathaway.

Just a few days earlier, Reuters reported that users and even a Meta employee could create “flirty” chatbots using celebrities' likenesses. For example, a Meta employee supposedly created Taylor Swift “parody” bots.

Furthermore, the report suggests that Meta allowed users to create chatbots of child celebrities like 16-year-old Walker Scobell.

Users could then potentially create topless images of the male child star when prompting the chatbot. The AI avatar then responded, “Pretty cute, huh?”

In light of this, Meta states that it’s beginning to employ temporary changes that will be implemented permanently at a later date.

Meta’s spokesperson Otway told TechCrunch that it's likely that Meta’s AI userbase will decline due to these changes.

So, will Meta follow through with these changes?

Considering Meta has been poaching top AI talent and has spent hundreds of millions of dollars on its new employees to obtain a competitive edge, it’s possible that this policy change might not be in the company’s best interest.

However, Meta advertised its commitment to child safety by tightening teen message controls, expanding safety features on its apps like Instagram, and using AI to identify teen accounts to push them into more age-appropriate online environments.

Only time will tell whether Meta cares more about retaining its AI userbase or protecting children from inappropriate behaviour online.