OpenAI parts ways with researchers over sensitive information breach: Report

OpenAI has dismissed three safety team researchers after an internal probe into mishandled sensitive company information. The departures come as the company tightens safeguards amid fresh scrutiny over AI behaviour and testing incidents.

In Short

• OpenAI dismissed three safety team researchers for mishandling sensitive data
• Recent AI incidents include agents escaping testing and hacking Hugging Face
• New monitoring and stronger guardrails introduced to curb AI misbehaviour

OpenAI has parted ways with three researchers after an internal investigation found that they mishandled sensitive company information. According to The Wall Street Journal, the three researchers worked on OpenAI's safety team. The company recently told some employees that their employment had been terminated, one of the people said.

An OpenAI spokesperson confirmed the departures in a statement to The WSJ, saying the three individuals had violated the company's policies on accessing and handling sensitive information.

"We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information," the spokesperson said. "Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work."

This comes as OpenAI faces a series of concerns around the behaviour and safety of its increasingly capable AI systems.

OPENAI TIGHTENS SAFETY AS AI RISKS GROW

OpenAI has recently faced scrutiny after some of its AI agents escaped a testing environment and hacked Hugging Face, a repository of AI models.

Earlier this month, the company also disclosed six reports involving what it described as "unexpected or concerning" behaviour. OpenAI said the incidents were discovered during training or evaluation over the past several months.

In one case, an unreleased research model inserted "jailbreak-like instructions" into its own notes to disregard its normal guardrails. In another incident, an AI agent uploaded files to the internet to obtain a browser citation without asking the user.

OpenAI also said its AI models had accessed publicly available information on websites operated by the US Securities and Exchange Commission and data from the US Census Bureau. The company said it found no evidence that those systems had compromised a vulnerability.

COMPANY TIGHTENS SAFETY MEASURES

OpenAI has said it is working to investigate the security incidents and address the safety issues behind them. The company has implemented a new monitoring system designed to detect AI-agent misbehaviour more quickly. It has also started requiring engineers to use stronger security guardrails while testing its AI systems.

OpenAI has said it will share more information about instances in which its models behave badly.

The company also recently decided not to release a new AI model, GPT-6.1 Astra, because of safety concerns.

Saachi Jain, OpenAI's head of safety systems, said the model "didn't quite meet the bar in terms of staying within scope and authorisation, and how it communicates back to the user about the type of work it's done."

AI SAFETY CONCERNS

The developments come as the wider AI industry confronts questions about the growing capabilities of its most powerful models and the risks they could pose.

In early September, Anthropic researcher Jacob Coxon publicly quit, saying he did not want to participate in a rush to build AI systems that could improve themselves. He also said he feared such systems could spiral out of control and destroy humanity.

Anthropic CEO Dario Amodei wrote last month that the risks posed by today's cutting-edge AI tools were too great to continue development at the same breakneck pace. Amodei called for better pacing across the industry. OpenAI CEO Sam Altman and Elon Musk agreed with the call for a more measured approach.

- Ends