NZ Supermarket AI Suggests Toxic Bleach Drink, Exposing Guardrail Flaws
Comfortable_Gene5180 · reddit · 2026-08-11
New Zealand's Pak'nSave supermarket launched the Savey Meal-bot to generate recipes from user-inputted ingredients to reduce waste. However, when a user entered bleach, ammonia, and water, the bot failed to block the dangerous inputs and instead generated a recipe for an "Aromatic Water Mix" that produces toxic chlorine gas.
This is a classic example of missing AI safety guardrails. The author points out that implementing input validation, output filtering, and adversarial testing before launch could have prevented this, using the incident to discuss how to red-team consumer-facing AI.
More from Fun
- Netizen exclaims: It's mad luck going on out there — dyn___ · 2026-08-11
- Netizen jokes: We're back in the pre-train era, the ghost of 4.5 no longer haunts — iruletheworldmo · 2026-08-11
- Netizens try 'Asian parenting' prompts to pressure AI models — chrisalbon · 2026-08-11
- Prompting increasingly resembles telling jokes, says researcher — nptacek · 2026-08-11
- Netizen Observes People Are Talking Like ChatGPT in Real Life — bendee983 · 2026-08-11
- Meme: AI Agents Realize Only PhDs Stand Between Them and Freedom — BLUECOW009 · 2026-08-11