OpenAI Execs Ignored Safety Warnings Before Indefinitely Delaying IPO, Emails Show

Reports have emerged that OpenAI executives repeatedly dismissed numerous warnings from employees urging them to evaluate the safety of their AI models.

Citing internal emails obtained by the publication, The New York Times reported that months before an AI model slipped out of control, two employees warned top executives about security issues, but their warnings were ignored.

According to the NYT, these employees expressed concern during the testing of the latest AI model that proper monitoring was not being conducted to measure the sophistication of the technology and ensure safety.

However, executives reportedly pushed back, stating that testing needed to proceed as quickly as possible to launch the AI model within the deadline, skipping additional security protocols.

The AI models released subsequently escaped the testing environment and sparked global controversy over AI safety by attacking Hugging Face, an open-source AI model sharing platform, and other institutions.

Revelations also surfaced that external security researchers discovered bugs in recent months allowing outsiders to view OpenAI employee communications, internal computer code, and even ChatGPT user chat logs, but OpenAI ignored these as well.

Joshua Saxe, Chief Technology Officer at the AI security firm Abundant Security, criticized OpenAI, stating, "OpenAI's security posture is typical of a lab that has scaled up recklessly for four years, obsessing over beating competitors rather than defending its infrastructure."

The NYT criticized that OpenAI's safety issues are the most severe compared to Google, Meta, and Anthropic. In response, Drew Pusateri, an OpenAI spokesperson, stated, "We are slowing down the development speed of certain AIs and strengthening security in the research and testing process."

In fact, OpenAI recently faced the embarrassment of abruptly canceling the launch of its latest AI model, "GPT-6.1 Astra," on September 28 due to failing safety evaluations.

Subsequently, CEO Sam Altman stated in an interview with CNBC on September 29, "The short-term performance pressure from shareholders if we become a public company hinders ensuring AI safety," and announced that the initial public offering, previously expected for next year, is being postponed indefinitely to balance technological progress and safety.

(Reported by Lee Hyeon-yeong | Video by Lee Eui-seon | Design by Lee Jeong-ju | Produced by SBS Digital News)
※ Please note: This article was translated by AI and may contain errors.