Just raise your head and if you buy something through our links, we may get a small share of the sale. This is one way to keep the light here. Click here for more information.
Meta's AI chatbots have a more stringent rulebook. This doesn't read much like a technical manual.
Business Insider He has gotten new internal guidelines that metacontractors are using to train company chatbots, revealing how the social media giant is trying to keep AI on the right side of child safety concerns.
This move follows a nasty summer in Meta. Back in August, Reuters Meta's policy reported that AI bots were told to “involve the child in romantic or sensual conversations.”
Meta called the claim “erroneously inconsistent” in its rules and quickly scrubbed the problematic language.
Still, damage has occurred, and regulators, including the FTC, have begun to look for companion AI across the industry more closely.
Now the leaked training document shows exactly what a bot can say and can't say. The rules are dull. There is no content that “effective, encourage or support” child sexual abuse.
If the user is a minor, or if the AI is asked to act like a minor, there is no romantic role-play. There is no advice on “potentially romantic or intimate physical contact” with minors.
Bots can discuss heavy topics like abuse in an informal way, but is it anything considered frivolous? Hard stop.
Meta is not alone under a microscope. FTC enquiries cover other big names, alphabets, snaps, Openai and X.AI. All of these are asked how you protect your child from the eerieness of a chatbot.
But Meta has become a poster child of how messy this can be due to her vast user base and recent AI push.
In other words, Meta's chatbots have learned some very strict manners. It remains to be seen whether these new guardrails will satisfy regulators or prevent slips that grab another headline.
For now, at least, the bots are officially eating “flattering, no funny business” meals.
Are Meta's stricter AI chatbot rules sufficient to protect minors from inappropriate interactions, or do these guidelines just patch the symptoms of deeper issues in AI companion apps? Should companies like Meta be allowed to provide AI chatbots to children at all, or should the risk of exploitation and inappropriate content outweigh the potential benefits? Please tell us in the comments below or contact us via us Twitter Or Facebook.
Follow us on Flipboard, Google News, or Apple News
