Microsoft Warns Of A Vulnerability That Exposes Conversation Content In Chatbots Powered By Artificial Intelligence

The blog published by the company on November 7 stated that the vulnerability allows hackers to reveal sensitive details about instructions by analyzing the sizes of data packets and the timing of their transmission over the network, without the need for decryption.

Microsoft explained that attackers can classify instructions related to specific topics only by analyzing responses flowing from large language models, as tests showed that the attacker can achieve 100% accuracy in classifying targeted conversations, with his ability to capture between 5% and 50% of the total conversations related to a specific topic.

The company warned that the vulnerability raises significant privacy concerns, especially in repressive regimes, where discussing topics such as protests, elections, or prohibited topics may expose users to surveillance and targeting.

It also noted that a government agency or Internet service provider may be able to identify users who discuss specific sensitive topics such as money laundering or political dissent.

In response to this vulnerability, several companies, including OpenAI, Mistral, Microsoft, and AIX, have deployed protection measures. OpenAI and Microsoft Azure have provided a mechanism that adds a random string of texts of varying length to each response in order to hide the true length of the code sentences associated with the instructions.

Microsoft advised users to avoid discussing highly sensitive topics when using chatbots over untrusted networks, to use virtual private networks (VPN) to secure the connection, and to stay informed of providers’ security practices.

It is noteworthy that this vulnerability comes in the wake of researchers revealing a new set of vulnerabilities that affect “GPT Chat,” which may allow attackers to steal personal information from users’ memory and chat logs.

Source: Agencies

Stealing people’s talents through artificial intelligence

A sinister plan to weaken copyright law and allow tech companies to free ride on people’s talent. This is a scandal, and the nation will suffer the consequences. Andrew Lloyd Webber – The Guardian