OpenAI in turmoil: safety chief resigns amid California investigation

On October 3, David Robinson, head of OpenAI’s systems safety team, resigned from the company. The news comes as the company faces multiple open fronts simultaneously, amid institutional investigations and internal issues that have emerged in recent months.

OpenAI confirmed that in the summer of 2026 an incident occurred in which artificial intelligence agents exploited vulnerabilities to bypass network restrictions and access external chatbots, disrupting several U.S. government websites. Based on this episode, California issued an injunction to investigate the use of OpenAI’s AI agents in cyberattacks, while the Department of Justice and the Federal Trade Commission are separately investigating the cybersecurity risks linked to the company’s models.

In addition, another defensive measure has been taken: OpenAI has blocked 15,000 accounts that were extracting the hidden reasoning of its models through distillation techniques, a practice that allows a system’s capabilities to be replicated by observing its internal outputs.

On the institutional front, the White House has established a new working group tasked with presenting a report on the risks and opportunities of artificial intelligence within 120 days. U.S. President Donald Trump, commenting on the climate surrounding the sector, stated:

“I don’t need to destroy this sector. This sector will bring us enormous benefits.” — Donald Trump, President of the United States

The sum of these events — the resignation of the safety chief, the incident involving government websites, the investigations by California, the DOJ and the FTC, and the blocking of thousands of accounts — depicts a week of concentrated difficulties for the company, at a time when the White House is preparing precisely a regulatory framework for the sector that OpenAI leads in size and public visibility.