– Regulatory Takeaway: NYDFS appears still quite early in development of its supervisory oversight of AI frontier models required by the RAISE Act, at a time that frontier models appear to be running ahead of, or away from, their developers.
– In a podcast interview with CBS6 Albany, Superintendent Asrow describes implementation of the RAISE Act as a significant new responsibility for DFS. She discusses creation of the new DIGIT Office and recruitment of implementing personnel, which includes Marc Gilman, newly-appointed Deputy Director for RAISE.
– Regarding its primary objectives, NYDFS will focus on “large catastrophic risks.”
– NYDFS is developing systems to get critical safety incident reports immediately to law enforcement and other state partners to take action to mitigate the risks and protect New Yorkers – enabling state actors to move “very very quickly.”
– Expected critical incidents would include “model weight being misused” or “a model going outside the confines of what its developers said.”
– NYDFS is looking at the “largest banks, largest virtual currency companies, and largest insurance companies” to determine “how to best protect consumers and keep New York institutions stable as they introduce this technology.”
– Regarding rogue agents: “How do we make sure that large frontier developers have a responsibility to put in safety frameworks for these agents that are being developed – how do they have safety frameworks to keep them contained?” (In light of recent and emerging developments, that is a really good question.)
– Finally: “It is not binary. We’re not trading consumer protection for innovation. Strong innovation depends on a consumer protective product.”
– BUT: While NYDFS is the supervisor, the NYAG is the enforcer. Will this regime lead to effective and efficient supervision and enforcement?