Altman releases OpenAI's five principles, hinting at future safety restrictions on model capabilities

Crypto World News: OpenAI CEO Sam Altman personally signed and published five operating principles for the company. In the article, he previewed that in some future periods, OpenAI may tighten users’ access to model capabilities for safety reasons, saying that there may be a time when it’s necessary to “sacrifice some empowerment to gain greater resilience.” Altman noted that key AI decisions should be made through democratic processes, rather than being decided solely by AI labs. He defended OpenAI’s recent practices such as large-scale purchasing of computing power, building data centers worldwide, and vertical integration, stating that the goal is to bring AI costs down to a level that everyone can afford, and that governments may need to explore new economic models to ensure value is shared. He also mentioned a specific risk scenario: extremely powerful models could lower the threshold for manufacturing new pathogens, requiring responses across the entire society. Altman looked back at the hesitation at the time to publicly release the GPT-2 weights, saying that in retrospect it was overly cautious, but it gave rise to OpenAI’s iterative deployment strategy that is still followed to this day.