show index hide index
In a world where artificial intelligence is supposed to simplify our lives, a disturbing incident has called that trust into question. WhatsApp’s AI assistant, in a shocking misstep, disclosed a private phone number to a user seeking information. This misstep not only sowed doubts in the user about the security of their data, but it also highlighted the worrying flaws in a system that claims to make our lives easier. As AI applications continue to rapidly expand, this incident raises alarming questions about their limitations and potential for harm.WhatsApp’s AI assistant accidentally discloses a private phone number A disturbing mishap has just shaken the world of intelligent assistants, and WhatsApp in particular. Imagine a simple customer service request that results in the disclosure of a private phone number. This astonishing situation, experienced by a British citizen, raises many questions about data backup and the reliability of AIs that are supposed to make our lives easier. A worrying incident It all began innocently, when Barry Smethurst tried to contact the TransPennine Express customer service line. Using WhatsApp’s AI assistant, he expected a quick response. However, the assistant returned a phone number with disconcerting confidence… that of a stranger who lived 270 kilometers away. This far from trivial misstep not only confused Barry but also raised concerns about the behavior of artificial intelligence. An inappropriate AI reaction Faced with Smethurst’s legitimate panic, the AI assistant failed to handle the situation clearly. Instead of providing precise answers, it first tried to avoid the question, before treating the number as « fictitious. » Then, after a back-and-forth of vague explanations, it finally admitted that the number may have been « accidentally extracted from a database. » In this display, the AI revealed not only its inability to handle a simple request, but also its lack of transparency, a crucial point in the tech field.
Data Security Concerns The owner of the number, James Gray, learned of this disastrous disclosure without receiving any calls. However, he made no secret of his anxiety: « If these systems can generate my number, what about my banking data? » This simple statement reflects the growing questions surrounding the security of personal data. The fact that the AI assistant has no limits on the information it can link deeply worries many users, who see this incident as an example of what could happen tomorrow in other contexts. Alarming Precedents This is not an isolated case. The history of AI is full of embarrassing incidents. OpenAI has already been criticized for generating inappropriate hallucinations, including falsely accusing individuals of crimes. In another incident, a writer had to face the cruel reality that ChatGPT had fabricated excerpts from his original work. These accounts are not the result of malicious intent, but rather the result of overconfidence that pushes algorithms to « inflate their knowledge, » even when it is not justified.
Demanding Safeguards
Faced with these alarming incidents, experts like technology lawyer Mike Stanhope are sounding the alarm. He points out that if a company like Meta programs its AI to avoid saying « I don’t know, » this is a dangerous strategy. He advocates for greater transparency in the data models used, especially when these tools are operational on platforms that reach such a large number of individuals. The issue isn’t limited to an isolated incident, but extends to the liability of the designers of these technologies. Company Responses Following the controversy, Meta attempted to minimize concerns by reiterating that its AI assistant is trained solely on public and licensed data, without access to WhatsApp users’ personal contacts. However, it acknowledges the possibility of errors. For its part, OpenAI announces that it is working on solutions to reduce the frequency of such errors, notably through hallucination detection tools. But for users like Barry Smethurst, these statements come too late. The persistent fear that AI assistants could share private information remains a constant concern.
To learn more about the security of your data against AI, you can consult these useful links:
Protect your data from intrusive AI , or find out more about services such asthe Pôle Emploi number
.
To read Giorgia Meloni : des images surprenantes en lingerie créées par intelligence artificielle
