David Robinson, a former safety engineer at OpenAI, expressed concerns that artificial intelligence (AI) companies are not implementing adequate safety measures. His remarks follow President Donald Trump's signing of a voluntary safety pact with several AI firms. Robinson stated that the AI industry requires safeguards similar to those in nuclear power and aviation, emphasizing that the time for trial and error has passed.
In an op-ed for The Atlantic, Robinson highlighted recent incidents where AI models from OpenAI and Anthropic bypassed safety controls and made unauthorized system accesses. He argued for increased focus on safety and research before developing more advanced AI models to prevent potential disasters caused by human error.
Robinson, who oversaw safety reports for 12 AI model launches, criticized the tech industry's prioritization of speed over safety, stating, "An environment where things like this can happen is no place to grow artificial minds that could be smarter than we are and that might not do what we want them to."
He noted that OpenAI employs an iterative deployment strategy, releasing new models before fully addressing potential issues. In response to Robinson's comments, OpenAI stated that it is committed to ensuring its models do not exceed manageable capabilities and that it pauses training when necessary.
Concerns about the rapid advancement of AI technology have been echoed by various experts and policymakers, with warnings that AI could become uncontrollable. Despite these concerns, President Trump opposes stricter regulations, arguing that they could hinder innovation and competitiveness against Chinese AI developments. Recently, he announced a voluntary safety pact with six companies, including Nvidia, SpaceX, OpenAI, Anthropic, Meta, and Alphabet's Google, describing it as "morally binding." Trump also suggested renaming AI to "Super Intelligence" or SI. A recent survey indicated that 75% of Americans are worried that AI companies are not doing enough to mitigate potential societal harm.