It would be too easy if I say that this is just another “AI made a mistake” story..!!

We all have interacted with chatbots. Those at times wasted our time. But the new AI chatbot (“AI-powered” if I quote as it is a magic word nowadays 🙂) can do something more than just wasting our time.. Read on.

We all remember the old support bot. They used to have this special skill of misunderstanding our query in seven different ways and still end with:
“was this helpful?”!! At worst, we lost patience or got annoyed but would not lose our account!!

I happened to come across these stories associated with Instagram while surfing.

The Verge

Reuters

So in June 2026, reports claim that attackers manipulated Meta’s AI-powered Instagram support chatbot during account recovery. The issue allowed password reset links to be sent to email addresses that were not linked to the actual account owners. The Verge mentioned that Meta confirmed 20,225 Instagram accounts were likely affected, and Reuters reported that high profile accounts were hit, including the former Obama White House page, Sephora and a senior U.S. Space Force official!!

This is where the story becomes less about Meta and more about every technology function rushing AI into operations. There is a big difference between a bot that explains a process and a bot that can “move” the process.

Because once AI can reset access, update records, trigger workflows, approve requests or send something on behalf of the organisation, it is definitely no longer “just support chatbots.”

If the journey involves access, identity, money, personal data or business critical action, then the question cannot only be: “How quickly can AI help?”

But ask a boring, important question: “What exactly is this AI allowed to do?” OR say “Does it know when to stop being helpful?”

The new risk is different. Some AI support tools may do too much, too confidently, in places where confidence is not a control.

That is the real leadership thought process we got to embed before we put AI deeper into operations.

We spent years complaining that bots were useless. And in Meta, it turns out that the next problem may be a bot that is useful to the wrong person. 😉