ChatGPT may reveal business secrets

● Risks can be controlled if businesses implement appropriate security measures.

● Team8, a cybersecurity company from Israel, has drafted a report on this issue with feedback from information security directors of many businesses.

World - ChatGPT can reveal business secrets — Photo: Gabby Jones/Bloomberg.

Team8, an Israeli cybersecurity venture, said in a report that businesses currently using some artificial intelligence applications like ChatGPT are likely to face the risk of exposing confidential customer information and their trade secrets.

The report, provided to Bloomberg News ahead of its publication, said the growing use of AI in chatbots and text editors could expose many businesses to data breaches and lawsuits. The report raised concerns that chatbots could be used by hackers to mine sensitive corporate information or conduct other damaging activities against the company. There are also concerns that confidential information currently provided to chatbots could be used by AI companies in the future.

Many big tech companies like Microsoft Corp. and Alphabet Inc. are racing to integrate artificial intelligence to improve their chatbots and search engines, training their artificial intelligence models on data sets mined from the Internet to give users a multi-functional tool that can answer any question they ask. According to the report, if these applications are provided with confidential or private data, it will be extremely difficult to remove the information provided.

“The application of artificial intelligence in the enterprise environment can lead to the risk of sensitive information, intellectual property, source code, trade secrets and other data being exploited and processed through data received directly from users or through APIs, including user privacy and confidential information,” the report said. The report rated these risks as “high” and said that these risks are “manageable” if appropriate cybersecurity measures are implemented.

Team8's report emphasizes that chatbot input is not provided to companies for training artificial intelligence, contrary to recent reports that input can be read by others.

“As of this writing, AI models trained on large-scale language data are not capable of updating themselves in real time, and therefore are not capable of using input from one person to respond to another, so concerns about that possibility are not warranted. However, future versions of these models are not necessarily going to have similar characteristics.”

The report also rated three other issues with AI integration as “high risk” and highlighted the growing threat of information sharing with third-party applications. Microsoft has integrated several chatbot features into its Bing search engine and Microsoft 365 software suite.

“For example, on the user side, third-party applications that leverage the APIs of artificial intelligence models can be compromised and granted access to the user's email and web browser, allowing the intruder to perform activities on that user's behalf.”

The report also assessed a “medium risk” that the use of artificial intelligence could increase discrimination, damage a company’s reputation or expose it to legal action around copyright issues.

Ann Johnson, corporate vice president at Microsoft, contributed to the report. Microsoft has invested billions of dollars in OpenAI, the company that developed the ChatGPT application.

“Microsoft encourages transparent discussion about evolving cyber risks within the cybersecurity and artificial intelligence communities,” a Microsoft spokesperson said.

Several chief information security officers from a number of US companies also contributed to the report. The Team8 report was also endorsed by Michael Rogers, former chairman of the US National Security Agency and US Cyber Command.

Nguyen Quang Minh (according to Bloomberg)

Source