 |
Users' phone numbers can be revealed by AI even without their permission. Photo: Deposit . |
This is one of the notable examples cited in Eileen Guo's MIT Technology Review article about the situation where AI applications arbitrarily disclose personal information, including phone numbers and home addresses, of many people without their permission or even their knowledge.
A similar case emerged on the Reddit forum. Users complained, seeking community help, about constantly receiving calls from strangers. They contacted lawyers, product designers, and locksmiths. All said they obtained the phone numbers through AI instructions.
In April, a doctoral student at the University of Washington reported accidentally obtaining a colleague's personal cell phone number while browsing on Gemini.
Where does AI get phone numbers from?
Chatbots like ChatGPT are trained on massive amounts of data, much of it from the internet. Therefore, it's possible that a website containing personal information, such as a random forum post, could appear in the chatbot's dataset and be included in the query results.
Even if the phone number isn't in the training data, the chatbot is capable of searching the web. These models scan large numbers of websites to return results for a request. If user information is found, the AI is likely to provide it automatically.
The deeper problem is that our information is everywhere on the internet, even without our knowledge. It's in articles published years ago, public service websites containing public records… It often doesn't appear in regular search results.
However, AI has the ability to deeply analyze all web search results, finding lesser-known results and providing them, potentially exposing users' personal information.
 |
Users may frequently receive calls from strangers because their phone numbers are leaked through AI chatbots. Photo: CBS News. |
As Guo explained, most chatbots have measures in place to protect personal data. For example, if a user directly asks ChatGPT for their phone number, the chatbot might say it cannot provide it due to security regulations.
However, when changing the query to search for a specific person's phone number, the AI is likely to provide it.
An example in this case is technology editor Jake Peterson of the website Lifehacker . ChatGPT provided him with two phone numbers of "Jake Peterson" that were publicly listed on the internet.
He received a similar response from Gemini when he directly asked for personal contact information, but Google's AI chatbot readily shared Peterson's X social media account, a sign of inconsistency in the AI's safety policy.
Is it possible to remove phone numbers from AI?
Users have limited options to protect their privacy from AI tools. Some have feedback policies, but the final decision rests with the developer.
OpenAI has a portal that allows people to request the removal of their personal information from ChatGPT responses, but the company reserves the right to refuse requests for various reasons.
Anthropic only provides supporting documentation explaining how they use user information, while Google allows users to request to opt out of processing personal data. However, this depends on the jurisdiction, such as in the EU and the UK, where regulations on personal data protection are strict.
According to Jake Peterson of Lifehacker , the most practical way is for each user to remove their personal information from the internet themselves. Some places, like California (USA), have data brokerage companies that allow users to request the deletion of their data.
You can also learn about some personal data deletion tools, such as Incognito or DeleteMe. It's important to note that while these tools can remove information from some places on the internet, users can't do much if AI companies have access to the data in their datasets.
Source: https://znews.vn/nguy-co-lo-so-dien-thoai-qua-chatgpt-post1652718.html
Comment (0)