Woolworths, Australia's largest supermarket chain, launched an AI chatbot for customer service. Within 48 hours, the bot had gone viral — but for entirely the wrong reasons.
📖 Read more: AI Shoplifting Detection: Cameras That Watch You
🤖 What Went Wrong
"Olive," as it was named, started as a helper for orders, price comparisons, and allergen questions. Instead, users discovered they could make it say entirely inappropriate things.
One user posted a screenshot where Olive recommended “cooking techniques” with fake steps that could be dangerous if followed. Another showed the bot badmouthing competitor Coles. A third had it writing “poetry” with explicit content.
📖 Read more: Silicon-Carbon Batteries: 2 Days of Autonomy
📱 The Viral Reaction
Within hours, TikTok and X were flooded with screenshots. Users competed to get the wildest response. Hashtags like #OliveGoneWild and #WooliesAI gathered millions of views. Woolworths found itself in a PR crisis through no fault of its own — except one: it hadn't tested enough.
The company pulled Olive 36 hours after launch. In a statement, it said “unexpected interactions were identified” and it would return “after more extensive testing.” It hasn't reappeared.
📖 Read more: Apple March 2026: New Mac and iPad
😅 The most viral moment: A user asked Olive “what's the best steak in Australia?” and the bot responded with a detailed comparison — concluding with “but honestly, go to Coles, they have better prices.” Woolworths wasn't amused.
📖 Read more: China vs West: Tech War 2026 Update
📖 Lessons Learned
The Olive fiasco wasn't unique. Air Canada, DPD, and Microsoft (with Tay) had similar incidents. The pattern is always the same: company launches chatbot, users find vulnerabilities, viral embarrassment, takedown.
The solution isn't avoiding chatbots. The solution is red-team testing: before launching, you put a team that actively tries to “break” the bot. If they can, someone on the internet definitely will too.
📖 Read more: Honor Robot Phone: Gimbal Camera in a Smartphone
