During a recent Salesforce event held in San Francisco on September 15, OpenAI CEO Sam Altman discussed the approach to managing risks in artificial intelligence development. He stated his belief that the industry can maintain safety standards through self-regulation without causing significant harm to the public.
Altman emphasized that safety in AI requires more than just verbal promises from companies. He advocates for a comprehensive approach that includes simultaneous improvements in model capabilities, alignment, security, and confidentiality. OpenAI has already taken practical steps by delaying certain frontier reinforcement learning training sessions.
The company has redirected a portion of its computing power toward monitoring, safety guardrails, and risk assessment protocols. Altman suggests that by fostering industry collaboration, conducting independent evaluations, and implementing stricter internal safety processes, companies can establish effective safeguards before government regulations are fully enacted.
