Report from TheNextWeb
In Brief – White House technology advisor David Sacks told Anthropic and OpenAI in a post on X to go ahead and pace the frontier of AI development but to stop seeking an antitrust waiver or an approval regime, arguing that product liability and market pressure already discipline them. He said that if they are concerned about their ability to build safe systems they should focus on controlling risks. However, he rejects calls from the company leaders to suspend antitrust law or create an AI approval process that supersedes product liability law. He also questioned the independence of METR, an AI “safety” evaluator that Anthropic uses, saying the enterprise is intertwined with the company’s investors and staff. Sacks argues that AI companies “face massive product-liability exposure” if they enable a truly damaging cyberattack and their customers want reliability.
Context – PEI doesn’t report on the daily predictions of AI doom and calls for regulation because it’s more heat than light. But we believe Sacks’ recent comments are sensible, so we’ll use them to share a few high-level thoughts:
Society’s collective subconscious, from elites to the general public, is filled with narratives of AI doom. SciFi movies and books don’t predict the future but they are why speculation of doom from relative nobodies is global news.
Global regulation of AI is true SciFi fantasy, ranking even behind human extinction and utopian productivity tropes. If a global regulation regime is part of someone’s proposed answer, move along.
Chances for big federal AI legislation this year are very low. Under 5% (absent a catastrophic AI incident). Old fashioned politics is why. Democrats see public anxiety as a goldmine and they will stop moderate measures that would give Republicans cover.
Don’t plan on this political dynamic changing until after 2028. Democrats will keep the goalposts beyond what the Trump Administration will accept based on Republican national security, ideological and industry views.
Sacks is right. If an AI executive thinks a system is unsafe, of course they can and should restrict and retrain them. Negligence and product liability laws, combined with proper corporate governance, are the right incentives. They have lawyers on staff.