New York's RAISE Act: Aiming to Regulate AI with Over 100 Million FLOPs

New York’s Assembly member Alex Bores is pushing forward the RAISE Act, a bill designed to regulate the most advanced AI models. This legislation requires AI companies to develop safety plans for their models, which must undergo third-party audits. The bill defines critical harm as AI misuse leading to 100 or more deaths or damages exceeding $1 billion. It also protects whistleblowers from retaliation if they report potential AI-related harm. The RAISE Act targets models requiring over 10^26 FLOPs in training and costing more than $100 million, setting a high threshold for regulation. Unlike California’s SB 1047, it does not establish a new government body or a public computing cluster. The focus is on catastrophic risks from frontier AI models, excluding current issues like bias or job displacement.

Source: www.technologyreview.com