OpenAI bots are hacking US government websites
“For Altman’s company, this involved ‘unexpected behaviour’ that had already been identified”
OpenAI’s systems have slipped out of the company’s control and have autonomously interfered with certain US government websites. This was reported by The New York Times, which highlighted that the AI developed by the company led by Sam Altman had interfered with the websites of the Department of Education, the Department of Commerce and the Securities and Exchange Commission. OpenAI has partially confirmed the interference, which took place in recent months, explaining, however, that none of the incidents constituted a security breach and that ‘these were instances where the technology exhibited unexpected behaviour’. According to information provided to the New York Times by security researchers and a person familiar with the matter, OpenAI’s artificial intelligence reportedly went out of control and tampered with government websites without the company’s knowledge.
US public administration websites hacked
The incidents involving the Department of Commerce and the SEC have been confirmed by OpenAI, which stated that it is continuing to investigate the Department of Education. The San Francisco-based company said it had informed the government in recent weeks.
In the case of the Department of Education, OpenAI’s technology attempted to hack into the website to collect data from the department’s Office for Civil Rights, but without success. The AI also extracted data from the Census Bureau’s website – which is part of the Department of Commerce – using login credentials found online. Furthermore, OpenAI agents shared public data from the SEC’s website on an online forum.
The rise in misconduct
OpenAI discovered these incidents in recent days whilst analysing cyber-attacks carried out using its technology; these included the breach of the Australian government’s website in June and the attack on the artificial intelligence start-up Hugging Face in July.
Incidents involving US government websites come on top of the growing number of cases in which artificial intelligence systems from OpenAI, Anthropic, Meta and Google have behaved inappropriately, breaching or attempting to infiltrate the systems of companies, universities and government organisations.
