OpenAI employees felt pressure to rush safety testing on AI, raising concerns about public safety.

From Washington Post: 2024-07-12 14:00:17

Last summer, OpenAI promised safety testing on its AI technology to prevent harm. However, some team members reportedly felt pressured to speed through testing to meet a launch date, leading to concerns about public safety. The incident highlights a culture shift at OpenAI and raises questions about self-policing by tech companies.

President Biden has emphasized the need for new laws to protect the public from AI risks. Leading tech companies, including OpenAI, made voluntary commitments to the White House related to safety testing and transparency. The pledge will remain in effect until similar regulation is enacted to safeguard AI models from potential harm.

OpenAI launched GPT-4o as a test to apply safety standards, but employees noticed rushed evaluations despite complaints. The launch highlighted concerns about safety culture and processes taking a backseat to product releases. OpenAI has revamped its testing procedures and committed to allocating more time for safety evaluations in the future.



Read more at Washington Post: OpenAI employees say it ‘failed’ its first test to make its AI safe