[DigitalToday reporter Yoonseo Lee (이윤서)] Generative artificial intelligence (AI) is being used not only for search and work assistance but also for everyday judgments such as whether to eat food. Google’s Gemini recommended discarding a chicken salad made 8 days earlier, citing food safety guidelines, when asked if it was safe to eat.

Tech site TechRadar reported on Aug. 10 that a user, via Gemini Live, pointed the camera at an opened chicken salad container and asked whether food made on Aug. 1 was still OK.

Rather than immediately deciding whether it was safe to eat, Gemini first provided guidance related to medical advice. It said the information it provides does not replace professional medical advice, diagnosis or treatment, and explained that questions related to a person’s health condition should be directed to a doctor or qualified medical staff.

It then recommended throwing the chicken salad away based on general food safety guidelines. Gemini said the recommended refrigerated storage period for chicken salad is generally 3 to 4 days, and judged that eating food made on Aug. 1 on Aug. 9 might not be safe.

Following Gemini’s recommendation, the user threw away the chicken salad and had a peanut butter and jelly sandwich for lunch instead. Gemini also replied that this was a safer choice than the chicken salad originally considered.

The case shows the expanding range in which generative AI is used as a reference tool for decision-making. The user had been using Gemini to pick clothing color combinations and this time broadened the scope of questions to food safety, which can be hard to judge by smell or appearance alone.

It also showed that AI can provide relatively conservative answers to health and safety questions. Gemini did not directly verify whether the food was actually contaminated, but assessed potential risk based on storage time and general food safety standards. Since a photo cannot confirm bacterial growth or actual spoilage, it chose to recommend disposal rather than consumption.

More users are also turning to AI to obtain health information. A Gallup survey found that more than 25 percent of Americans have used AI to obtain health information or advice. More than half of those surveyed said they asked AI first for health-related advice or analysis before visiting a hospital.

Still, generative AI responses have the limitation of being reference information only. AI cannot diagnose a user’s actual health condition or test whether food is contaminated. For that reason, Gemini also stressed in this question that it cannot replace the judgment of medical professionals while presenting general safety standards.

As the use of generative AI expands beyond search and productivity into everyday decisions, the need is being highlighted to recognize not only convenience but also the limits of answers when asking health and safety questions.