What happened: Well, Reddit’s new AI chatbot is already causing chaos. The bot, called “Answers,” is supposed to help people by summarising information from old posts. According to a report by 404media, the latest chatbot by Reddit is now suggesting hard narcotics to deal with chronic pain.
- When someone asked it about dealing with chronic pain, it pointed to a user comment that said, “Heroin, ironically, has saved my life in those instances.”
- And it gets worse. Someone else asked it a question, and it recommended kratom, an herbal extract that’s banned in some places and has been linked to some serious health problems.
- The scariest part is that Reddit is testing this bot right inside of active conversations, so its terrible advice is showing up where everyone can see it. The people in charge of these communities, the moderators, said they couldn’t even turn the feature off.
Why is this important: This is a perfect, if terrifying, example of one of the biggest problems with AI right now.
- The bot doesn’t actually understand anything. It’s just really good at finding and repeating what humans have written online – and it can’t tell the difference between a helpful tip, a sarcastic joke, or a genuinely dangerous suggestion.
- The real danger is that it’s putting this stuff right into the middle of conversations where real, vulnerable people are looking for help, and it presents the information as if it’s a fact.
Why should I care: So why does this matter to you? Because it shows just how risky these AI tools can be when they’re let loose without a leash, especially when it comes to something as serious as your health. Even if you’d never ask a chatbot for medical advice, its bad suggestions can start to pollute online communities, making it harder to know who or what to trust.
What’s next: After people (rightfully) freaked out, Reddit confirmed they’re pulling the bot from any health-related discussions. But they’ve been pretty quiet about whether they’re putting any real safety filters on the bot itself. So, for now, the problem is patched, not necessarily solved.
