00:00It's Benzinga bringing Wall Street to Main Street
00:02OpenAI on Monday proposed new safety standards for frontier artificial intelligence, with
00:06a focus on alignment research and recursive self-improvement, according to CNBC.
00:11The company called for international cooperation and standards covering frontier AI models,
00:15developers, and automated AI researchers.
00:18OpenAI said fully autonomous recursive self-improvement is not happening today and should not be
00:23pursued unless it can be done safely.
00:25The company warned that poorly controlled RSI could leave humans unable to oversee AI development.
00:31OpenAI cited the Hugging Face Agent hack as a preview of risks that could become more severe
00:36without safeguards.
00:37The proposals follow similar calls from Anthropic for stronger safety measures and
00:41independent evaluation of frontier AI systems.
00:43For all things money, visit Benzinga.com.
Comments