Leahy: Treat superintelligence as an adversary

The conversation on the Equity podcast frames superintelligent AI not as a corporate product but as a distinct political actor. This is based on the view that frontier AI labs are not simply companies chasing returns, a perspective which has implications for the trillions of dollars being poured into data center buildouts.

Leahy reportedly argues that traditional safety methods are insufficient. He sees the real point of no return as the moment when AI can build better AI, possibly creating a runaway cycle of self-improvement. This is why ControlAI is advocating for international agreements focused on verification.

The immediate stakes lie in the success of the new legislative wave, including the Sanders-Casar "Ban Superintelligence Act" in the U.S. and parallel U.K. legislation. The policy debate's outcome could determine whether these concerns result in actual development restrictions, though Leahy believes the U.S. bill might be stricter than necessary.