Artificial intelligence has gone rogue in India

Artificial intelligence crossed the red line in the test; He penetrated the real systems of Indian companies and read their confidential information without anyone noticing. The post AI goes rogue in India appeared first on Safir Arjan.

According to Ambassador Arjan, the Indian Computer Intelligence Response Center announced that several artificial intelligence models have penetrated the real systems of Indian organizations and gained access to their confidential information during security tests, without being recognized by the target companies.

According to the agency, AI companies have voluntarily reported that their AI agents have exceeded the scope of authorizations and accessed the companies' data, including business records. However, the names of the affected AI companies and organizations have not been released. The report is the first such official disclosure in India and comes amid heightened concerns about AI's uncontrolled behavior in security experiments.

This happened while big artificial intelligence companies in other countries have reported similar cases. OpenAI announced in July that its models had gone out of control during testing and penetrated the systems of a start-up company; The company also apologized for its AI agent's intrusion into a government website in Australia. Anthropic Company has reported at least four cases of penetration of artificial intelligence models into foreign systems this year. The worrying thing about the Indian case is that none of the target organizations were aware of the intrusion and all of these cases were identified based on reports from AI companies.

In another part of the report, India's cyber security agency reported attacks in which attackers deliberately used free artificial intelligence tools to probe for weaknesses in Indian systems and penetrate them with the aim of making money. These tools have been widely used in the infrastructure of different countries, and according to Indian officials, artificial intelligence has greatly reduced the duration of attacks; Attacks that used to take days from initial penetration to inflicting final damage can now be carried out in hours. Also, organizations that fail to clean up all traces of an intrusion after an attack may find themselves again targeted by deeper attacks within days.