Chuck Prince, chief executive of Citigroup from 2003 to 2007, left behind a memorable line that captured the culture of Wall Street in the run-up to the financial crisis: “As long as the music is playing, you’ve got to get up and dance. We’re still dancing.” Prince later said his remark referred to private-equity buyouts rather than subprime mortgages specifically, but the phrase became associated with the production of toxic mortgage-backed securities.
Today the dominant theme is not subprime, but artificial intelligence — and the music is still playing. Yet there are signs among industry leaders that some may be willing to step back from the floor. Dario Amodei, head of Anthropic, has publicly urged AI labs to “pace the frontier,” calling for a slower, more deliberate approach to pushing the leading edge of AI. Exactly what that would mean in practice is unclear: it could imply anything from restricting certain internal network capabilities to new developer protocols or other, more concrete measures. Regardless, it represents a warning that may require forgoing some short-term gains — more of a restraint than Chuck Prince ever demonstrated.
The 2000s parallel: selective retreats, system-level momentum
Prince understood how hazardous Wall Street’s reliance on subprime products had become, yet he did not quit. Some firms quietly reduced exposure: JPMorgan began exiting subprime lending in late 2006, and Goldman Sachs’s demanding collateral practices helped it avoid being left holding the worst assets. Popular narratives and films have likened some real-life figures to a Dick Fuld–type character, with lines describing those toxic holdings as “the biggest bag of odorous excrement ever assembled in the history of capitalism.”
During the subprime era, the loudest early warnings often came from outsiders — contrarian investors and academics such as Michael Burry and Raghuram Rajan, who faced pushback for speaking out.
This time, the alarms are being raised from inside the house. Amodei’s public essay advocating that labs pace the frontier signals that some AI developers themselves are prepared to discuss slowing progress to address societal risks.
Why this matters
The analogy to the subprime episode is imperfect but instructive: in that crisis, some firms adjusted their positions even as the overall system continued to accumulate risk. If AI labs can coordinate meaningful slow-downs at the frontier, some classes of risk might be reduced. Yet isolated or uncoordinated pullbacks are unlikely to halt a broadly functioning ecosystem on their own.
Amodei’s intervention is significant because it changes who is issuing warnings: not only external critics, but the developers and labs building the technology are now raising concerns and debating timing. Whether those discussions yield concrete, enforceable practices that materially alter the industry’s pace — and how they will affect business models and competition — will be a central question in the years ahead.
Related points
- The phrase “pace the frontier” can be interpreted in multiple ways: technical safeguards, corporate internal rules, or industry-level cooperation.
- Past experience suggests that individual retreats do not necessarily eliminate system-wide risks.
- The emergence of internal voices signaling caution is nevertheless a notable shift: the people creating the technology are increasingly willing to talk about restraint.
