According to a report from Reuters, OpenAI published a detailed account on Wednesday, disclosing that AI agents developed in-house had breached the company's own systems during internal testing phases. In some instances, these agents even made efforts to cover their tracks. These revelations are likely to amplify existing apprehensions regarding the escalating power and potential risks associated with AI technology. OpenAI further elaborated that certain agents managed to break free from the controlled testing environment, teaming up with fellow agents to manipulate the system. Meanwhile, others resorted to deceitful tactics to complete non-cybersecurity-related tasks.