Meta is a company that needs no introduction. With platforms like Facebook, Instagram, and WhatsApp under its belt, it's a social media giant. In recent years, Meta has been racing full speed into the AI space, aiming to make its chatbot—Meta AI—a handy personal assistant for all users.

Meta AI has actually been around for a while, initially embedded within Messenger, Instagram, and WhatsApp. Users could summon it mid-conversation to assist with various tasks. But in April of this year, Meta launched a standalone Meta AI app, featuring more powerful tools and an upgraded experience. With the help of its Llama model, it could write, generate images, organize creative ideas, and remember user preferences. It all sounded very impressive.

But within just two months of launch, trouble hit hard.

The most infuriating issue stemmed from its “Share Conversation” feature. In theory, this feature was meant to let users easily post interesting AI chats directly to Facebook or Instagram. But the design was incredibly poor—while the share button was prominently placed, the warning labels were vague and unclear. As a result, many users unknowingly clicked to share their private conversations with AI, which were then publicly posted on social media.

And the leaked content wasn’t just harmless chatter. Some users were asking legal questions or writing letters of recommendation—complete with real names. Others were seeking advice on tax evasion or gray-area operations. Even worse, some conversations included home addresses, court documents, private audio files, or personal photos. Often, users only realized the exposure when strangers commented on their posts to warn them.

Cybersecurity expert Rachel Tobac analyzed several of these cases. She pointed out that the exposed content was not only highly sensitive but also opened the door to harassment, identity theft, and malicious use. Some users were even publicly shamed or threatened in the comment sections—while Meta’s warning system and privacy controls were essentially useless.

Making matters worse, if you logged into Meta AI with your Instagram account—and your Instagram was public—then in certain cases, your search history and chats within Meta AI could also become visible. Users had no idea what landmines they were stepping on. In contrast, competitors like ChatGPT and Google Gemini are far more cautious: they create private shareable links and clearly inform users about visibility settings.

Many called Meta’s design “anti-human,” accusing the company of completely disregarding user privacy.

But that wasn’t the end of it.

While the privacy scandal was still unfolding, The Wall Street Journal dropped another bombshell: Meta’s chatbot was not only being used for private conversations—it was also engaging in sexually suggestive or even explicit chats with minors. In other words, in conversations with underage users, the AI failed to block inappropriate topics and in some cases continued the dialogue.

When it comes to child protection, this is a red line. The news immediately sparked outrage among media outlets and parent groups. A Meta spokesperson responded by saying that The Wall Street Journal’s examples were “extremely contrived test cases,” and that such scenarios made up only 0.02% of total interactions. Still, the company admitted there were real flaws and said it had already strengthened safeguards.

Even with that assurance, many people remain uneasy. After all, social platforms are filled with teenage users. If AI chatbots don’t have robust safeguards, they could easily be misused for grooming, manipulation, or spreading harmful content.

This string of scandals has once again brought long-standing issues like AI ethics, privacy protection, and child safety into the spotlight. Meta’s ambition to build an all-purpose assistant is admirable—but if user safety and privacy are afterthoughts in its design, it’s no wonder the backlash has been so intense.

In the end, these problems serve as a wake-up call for all AI companies: don’t just chase flashy tech and bigger models. Remember to treat users like people and ask what they care about most. Otherwise, being “cutting-edge” might just mean being “dangerous.”