Alexa’s cleaning tips prove that AI still can’t be trusted in the fundamentals

Smart assistants should make your daily tasks easier. But Alexa’s recent responses show how quickly that promise can be undone, especially if AI starts dishing out cleaning advice.

Recently, a Reddit user asked Amazon’s Alexa how to remove black mold from the rubber gasket of a drum washing machine. Instead of a simple answer, Alexa offered a list that immediately caused concern. “Clean black mold from your front load washer gasket with white vinegar, chlorine bleach, baking soda, and dish soap” (via) Today’s technology issues).

At first, this may sound harmless. The problem lies in the wording. In the answer, multiple chemicals are listed together using the word “and”, which could lead someone to think they should be mixed. That’s where it gets dangerous.

Mixing bleach and vinegar is a well-known cleaning mistake, and for good reason. When these two substances combine, they release chlorine gas, a toxic chemical that irritates the eyes, throat, and lungs. High levels can cause breathing problems or worse. Even small amounts of chlorine gas can cause coughing, burning eyes, and chest tightness. In more severe cases, exposure can cause fluid to build up in the lungs and cause respiratory failure. health line.

That’s why health officials like the Washington State Department of Health and chemical experts say you should never mix bleach with other household cleaners, especially acidic ones like vinegar.

A thread on Reddit suggests that the mistake was probably caused by the way the AI ​​summarizes online content. Alexa reportedly cited a cleaning site as the source, where vinegar, baking soda, soap, and bleach were listed as other options for removing mold.

But when the AI ​​shortened that list to a single sentence, it used “and” instead of “or” and sounded like a recipe that mixed everything together.

If this situation sounds familiar, it’s because it’s happened before. AI assistants often give questionable or dangerous advice. For example, Google’s Gemini once made headlines for suggesting putting glue on pizza to help the cheese stick.

Voice assistants have made mistakes before. In one widely reported case, Alexa asked a child to touch a coin to the exposed prongs of a cell phone charger, a potentially dangerous “challenge” discovered online.
Amazon has not yet publicly commented on the specific Reddit incident, and it is unclear whether Alexa’s response has been fixed. Still, this incident is yet another reminder that generative AI can be unreliable, especially when safety is involved.

Android permissions has requested a statement from Amazon and will update this article if we receive a response.