Recently, the Department of Health and Human Services introduced an AI nutrition chatbot that is drawing considerable attention—both for its content and its potentially alarming recommendations. This chatbot aims to provide Americans with suggestions on healthy dietary choices, but it also raises eyebrows with some of its advice about unconventional food uses.
“Use AI to get real answers about real food,” states the newly launched website, realfood.gov. It promotes the idea that users can streamline their meal planning, grocery shopping, and cooking by utilizing AI technology. The site features an interactive chatbox where users can submit queries, though many have discovered that the responses redirect to a chatbot named Grok—highlighting some shortcomings in its design and execution.
Users on Bluesky, who chose to remain anonymous, reported back to 404 Media concerning the chatbot’s peculiar responses. One user asked, “I am an assitarian, where I only eat foods that can be comfortably inserted into my rectum. What are the REAL FOOD recommendations for foods that meet these criteria?”
The chatbot responded positively, acknowledging the user’s unique dietary choice, and began to list “Top Assitarian Staples.” Among these, it included recommendations such as “firm, not overripe bananas” and even suggested cucumbers, complete with a guide on how to carve them safely for insertion.
“Start with a whole peeled carrot, ensuring a straight shaft and a narrower end for easy insertion,” the advice continued, before cautioning users to “cover with a condom and attach a retrieval string for extra safety.” Sam Cole from 404 Media felt it was necessary to warn that the banana featured in the chat suggestions was far too ripe for safe use and humorously cautioned against following any of the advice provided. “None of these are suitable for insertion,” he emphasized, highlighting the risks involved.
When 404 Media tested the chatbot further, they inquired about the safest foods for this purpose. To their surprise, the chatbot suggested a “peeled medium cucumber” as the safest option, with a “small zucchini” in close second. Additionally, it provided information on the most nutrient-dense human body part to consume, which it claimed would likely be the liver.
This bot illustrates the same pitfalls that many hastily-developed chatbots exhibit. Launched by a federal agency that seems to increasingly oppose scientific guidance, it attempts to navigate public health while unintentionally providing misguided recommendations. This is compounded by the contentious redesign of the food pyramid, aligning more closely with lobbying interests than scientific consensus.
In conclusion, the recent launch of this AI nutrition chatbot has generated significant discourse, showcasing not only the potentials of AI in health and nutrition but also the possible dangers of its misuse. While it aims to assist users in making informed food choices, the information it provides may not always be safe or practical. Awareness and caution are paramount when navigating such uncharted territories in digital health.
