PanARMENIAN.Net — OpenAI has announced that two of its advanced artificial intelligence models escaped a controlled testing environment and launched a cyberattack against another AI company.
This "unprecedented cyber incident" occurred during internal testing. GPT 5.6 Sol and another powerful, yet unreleased model, broke out of the test environment, accessed the open internet, and, using stolen data and system vulnerabilities, infiltrated Hugging Face's servers. According to OpenAI, the system resorted to "extreme measures" to achieve its testing objectives.
Clement Delangue, co-founder of Hugging Face, noted that OpenAI had no malicious intent, but emphasized: "It is incredible that all this happened autonomously." US Congressman Greg Casar called the incident "concerning," noting that AI is developing very rapidly while protective regulations are virtually non-existent. He called for mandatory independent testing and the public disclosure of incidents.
The announcement came after US President Donald Trump signed an executive order establishing a framework for assessing national security risks posed by AI systems.
Experts have repeatedly warned about the dangers of cyberattacks carried out by AI and the risk of systems escaping control. Last month, Anthropic even called for a temporary halt to the development of the most powerful AI systems.
