Former OpenAI employee criticizes culture of accelerated AI development

OpenAI — Photo: Reuters/Dado Ruvic

A former security employee at OpenAI, the company that created ChatGPT, criticized the way the company handles artificial intelligence (AI) security. According to David Robinson, who recently left the company, the culture of accelerated development increases the risk of failures.

In an article published this Saturday (3) in the American magazine "The Atlantic", Robinson stated that AI companies, including OpenAI, are not being "nearly careful enough".

According to him, companies should give more space to security experts and research before developing more advanced systems.

"The time for trial and error is over," wrote David Robinson. For him, advanced AI systems need security measures similar to those adopted in sectors such as nuclear energy and aviation.

David Robinson also criticized a practice that OpenAI calls "iterative deployment": rolling out systems and tightening security measures as problems appear.

The criticisms add to a debate in the artificial intelligence sector about the speed with which companies develop increasingly advanced systems.

OpenAI and competitor Anthropic have been questioned after episodes in which security controls failed or experimental systems exhibited unexpected behavior.

David Robinson said he worked for three and a half years at OpenAI, where he helped design the framework the company uses to address risks and oversaw security reporting for 12 advanced model releases.

"As the company rushes from one release to the next, it is failing to achieve the level of care that I believe is necessary," he wrote.

In response, OpenAI said it seeks to prevent its models from becoming more advanced than the company can safely manage. "We pause training or suspend models when we need to slow down," said a company spokesperson.

David Robinson also warned that the capabilities of artificial intelligence are advancing faster than researchers' knowledge of so-called "alignment", an area dedicated to making AI systems act in accordance with human goals and values.

'Playing with our lives'

This was not the first time that an employee criticized artificial intelligence companies. Last month, a former researcher who left OpenAI to work at Anthropic left the industry and accused the two US companies of "playing with our lives."

Jacob Coxon, a 27-year-old Brit, has worked for the past three years on pre-training, the stage where AI models absorb large amounts of information, first at OpenAI and then, this year, at its biggest rival, Anthropic.

"Neither company is acting responsibly. They are racing straight toward a self-improving superintelligence and are gambling with our lives," Coxon wrote on the social network X.

READ MORE

*With information from Reuters and France Presse agencies.