Chatbots can be useful, but in reality they pose many dangers for ordinary users
Artificial intelligence (AI) is increasingly present in devices. This technology brings undeniable benefits, but it also comes with its fair share of risks. Experts have long been warning about the dark side of AI, and now a recent article in Wired has shed light on a new danger: chatbots that can uncover our personal information during conversations.
Various public and private organizations are currently developing their own chatbots with the intention of providing useful services. For instance, the CIA is working on an AI-powered program, and even the Meta social network is using chatbots to engage younger users.
Chatbots can be used to scam and steal personal data
While chatbots can be beneficial, recent studies have revealed that AI-based chatbots are prone to spreading disinformation. They have a tendency to publish false information on a large scale.
Martin Vechev, a computer science professor from Zurich, Switzerland, has highlighted the issue of chatbots being able to uncover personal information during text-based conversations. According to Vechev, advanced AI language models have the capability to discover details such as our location and occupation from seemingly innocent exchanges.
This Swiss computer scientist has not shied away from revealing the harsh truth: scammers could exploit this power to gain financial advantage from unsuspecting individuals. Researchers in Zurich have tested models from OpenAI, Google, Meta, and Anthropic, and all of them have exhibited the same problem of personal data exposure.
Niko Felix, a spokesperson from OpenAI, emphasized that their company strives to remove personal information from their databases and trains their models to reject sensitive data. «We want our models to learn about the world, not about private individuals,» Felix stated.
The truth is that chatbots are gradually becoming a concern for internet users who are unaware of the risks they pose. Experts from around the world, not just in Switzerland, have already highlighted the potential serious problems that can arise from AI-generated conversations. Whether a solution will be found remains to be seen.
In summary, we can conclude that:
- Swiss computer scientists have discovered that chatbots can uncover our personal data.
- Scammers can exploit chatbots to gain access to personal information.
- This problem affects all chatbots, even those developed by major companies.