Yesterday afternoon, reporting from the New York Times indicated the White House was considering another executive order on AI, this time establishing an “AI working group” that could oversee pre-deployment testing of powerful AI systems. This could be a significant pivot. A full statement from FLI is below.

Key Context:

• Mythos was a wakeup call: the capabilities of advanced AI systems are not theoretical anymore. Anthropic’s latest model could have wreaked havoc on the internet, our economy, and undermined national security. The only reason it didn’t? Anthropic simply chose not to release it broadly. But it could have.
• The President takes note: The President vocalized some measure of support for a “kill switch” after it became clear that the cybersecurity capabilities of AI pose a serious threat to the financial system. Our statement on that is here.
• Congress is getting more serious: Senator Blackburn released a groundbreaking proposal for federal legislation on AI. The bill quickly earned broad support of groups across the ideological spectrum.

Anthony Aguirre, President and CEO of the Future of Life Institute, released the following statement in response to reports that the White House is considering an executive order establishing an AI working group:

“We are still waiting on details, but the premise is enormously encouraging. We need an ‘all of government’ approach to properly regulate advanced AI systems, which pose a threat to U.S. national security, global stability, and American society. The White House clearly understand the stakes.

Mythos was a wakeup call. Advanced AI systems are no longer just a commercial product, nor are the potential dangers hypothetical anymore. No single company should have the discretion to release AI systems that undermine our national security or wreak havoc on our economy. A whole-of-government working group — led by impartial experts — is necessary to help determine what kinds of pre-release evaluation and guardrails are required to keep Americans safe.

Our government must have the capacity to conduct thorough pre-deployment evaluation of the most powerful AI systems, particularly in a national security context. This kind of oversight is how we manage risk in aviation, in pharmaceuticals, in nuclear power, and in every other field where the cost of getting it wrong is high. This move by the White House is also a recognition that Big Tech cannot self-regulate, and that the status quo is unacceptable. We are glad to see the administration moving past the false choice between safety and innovation.

FLI is ready to work with the administration, Congress, and industry to ensure this AI working group produces concrete, enforceable measures, including a meaningful pre-release review process with real teeth. The power of the technology demands a commensurate response, and this news suggests that the White House understands that.”