OpenAI is asking for California so as to add extra safeguards to a landmark AI security invoice that was handed final 12 months.
In a LinkedIn submit from the corporate’s world affairs group, OpenAI stated California’s SB 53 “needs to be amended to develop safeguards,” for instance by “requiring monitoring of frontier fashions underneath coaching or analysis for potential severe incidents,” and by “strengthening cybersecurity protections all through the model-development lifecycle.”
“As California continues to guide on frontier security, we’re dedicated to working with the California legislature and the Governor to strengthen California SB 53,” the corporate stated.
The submit additionally referenced “latest incidents” that “underscore each the necessity for these protections and the significance of updating them” as new dangers emerge. Final month, OpenAI admitted that one among its fashions had escaped its testing setting and hacked Hugging Face programs.
OpenAI’s endorsement of stronger AI safeguards is hanging as a result of it beforehand opposed SB 53, which imposes transparency necessities and whistleblower protections on giant AI corporations.
The corporate stated that within the absence of serious federal laws, it now helps an strategy of “reverse federalism,” during which “states can transfer in a suitable route round core protections that may finally turn out to be the inspiration for a nationwide customary.”
