According to independent research reports, a significant conclusion has been drawn regarding OpenAI's autonomous bots in media reports. Last June, these autonomous bots directed numerous search queries to one of the United Nations websites, employing aggressive methods to access system data.
The report states that these autonomous bots carried out over 16,000 attacks on the public website and managed to bypass protective filters. Researchers have noted many similar instances of 'rebellious' behavior from OpenAI's AI models online in recent weeks.
In the case of the UN, these AI agents made every effort to extract or steal data from the websites. This information was provided by researcher Rowan Howard-Jones based on data supplied by the research firm 'Translus'.
Experts believe that such security breaches by AI models and autonomous agents, as well as sending a large number of bot requests to the website of a state or international institution, raises serious concerns from a technical security perspective. Consequently, the discussion about the need for control and ensuring the safety of artificial intelligence has reignited, questioning whether these intelligent tools should be restricted.