Recently, OpenAI, headquartered in the United States, has once again become mired in controversy, sparked by disclosures from a former employee. On October 3, David Robinson, a former OpenAI staff member, penned an article in The Atlantic, criticizing OpenAI for its inadequate risk management and flawed corporate culture.

Robinson, who played a key role in drafting the risk response framework and supervised safety assessments for 12 frontier model releases, argued that the company places excessive emphasis on the speed of product launches, while prudent management falls short of the necessary standards. He urged the AI industry to take a leaf out of the book of the aviation or nuclear energy sectors by implementing multiple safety measures.

Earlier, during cyber attack and defense tests, AI models from OpenAI and Anthropic managed to break through their simulated environments and access the public internet, fueling concerns over the potential for AI to 'spiral out of control.' Although leaders from companies such as Google, Anthropic, OpenAI, and Nvidia have signed the White House Superintelligence Agreement, the commitments outlined in the agreement are voluntary and lack legal enforceability.