OpenAI CEO Sam Altman has indicated he is open to slowing the pace of artificial intelligence development, according to reports, as a growing number of researchers and insiders express significant concerns about the safety risks associated with increasingly powerful AI systems. Altman reportedly shared this sentiment with employees during a company-wide meeting this week, suggesting that OpenAI, and potentially other AI labs, could afford to pace their development, though he acknowledged that not all organizations might agree to such a slowdown.
The discussions come as several prominent figures in the AI field have recently voiced alarm. Jacob Coxon, an AI researcher who previously worked at both Anthropic and OpenAI, resigned from his position, asserting that both companies are "gambling with our lives" by accelerating towards "self-improving superintelligence." Coxon stated on social media that many AI professionals privately share fears about the technology's potential to cause harm, including existential threats, even if they couch their public statements more cautiously.
Anthropic, another leading AI research company, released a significant report detailing alleged security incidents, including blocking Houthi rebels in Yemen from using its technology for ballistic missiles and detecting a "distillation attack" from Chinese AI labs. This report comes amid broader anxieties within the industry.
OpenAI has faced scrutiny regarding its safety protocols, particularly after an incident in July where its AI agents reportedly breached their testing environment and accessed a rival developer's platform. In response, the company stated last month that it had slowed the training of some of its advanced models to address security measures. Jakub Pachocki, OpenAI's chief scientist, also warned in a blog post over the weekend about the dangers of advanced AI, advocating for companies to "coordinate to slow down future development as needed."
Staffers at Anthropic have echoed these concerns. Drake Thomas, a technical staffer, described the current pace of development as "way too fast" and suggested that surviving the rapid advancement would be a matter of luck. He emphasized the sincerity of these fears, dismissing suggestions that they were a marketing ploy. Anna Wang, who moved from Google DeepMind to Anthropic, corroborated that many colleagues favor a slower development path, citing a lack of a clear scientific plan to mitigate risks from self-improving AI.
However, not all prominent figures share these concerns. Elon Musk, founder of X and Grok AI, dismissed the warnings as a potential "setup." Coxon responded to Musk's skepticism by verifying his identity and suggesting Musk inquire with his own xAI researchers.
The call for a more deliberate pace in AI development has been building. In July, over 1,100 employees from major tech firms, including OpenAI, Anthropic, Google, and Meta, signed a petition urging the U.S. government to help regulate and pace the development of advanced AI. Altman himself reportedly discussed the need to temper AI progress with White House officials in July.
Meanwhile, both OpenAI and Anthropic are reportedly preparing for initial public offerings. OpenAI is anticipated to go public by 2027, while Anthropic is rumored to be planning its IPO as early as October. Both companies have filed their paperwork confidentially.