AI’s leaders have a problem: They fear the race they’re trying to win

A pressing dilemma emerges in discussions surrounding artificial intelligence as leaders grapple with the potential risks their creations may pose to society. The debate intensifies when experts like former Anthropic and OpenAI researcher Jacob Coxon voice alarm, warning of a dangerous path toward self-improving superintelligence. Coxon joins other AI insiders in leaving their companies to sound the alarm, reflecting the complex "exit or voice" dilemma faced by those who believe their institutions are straying off course.

Economist Albert O. Hirschman's concept of "exit versus voice" highlights the tension between staying within an organization to effect change or leaving to speak out publicly. For those who remain, the challenge lies in balancing access and influence with the responsibility to safeguard the public interest. However, the timing of dissent becomes increasingly difficult when one has contributed to a system they later deem dangerous, as separating principled criticism from retrospective justification becomes increasingly complex.

Frontier lab executives confront a variant of this dilemma, as Anthropic CEO Dario Amodei calls for a slowdown in the advancement of powerful AI models to allow for adequate safety measures. However, achieving this requires industry-wide coordination, which poses a significant challenge given the competing commercial and geopolitical interests at play.

OpenAI CEO Sam Altman echoes these concerns, advocating for a more measured approach to AI development and acknowledging the possibility of AI surpassing human control. He also hints at the potential for collective action among leading AI companies to address these concerns.

The dilemma for these leaders shifts from whether to remain within their organizations to whether they should adopt a more cautious pace in their AI development. The question then arises: what does it mean to warn of potentially catastrophic consequences while simultaneously pursuing the most powerful AI? At some point, warnings must translate into concrete actions that reflect the recognized risks.

Altman now finds himself publicly navigating this decision-making process, grappling with who should set the limits and whether competing companies can agree to decelerate their development efforts collectively. Fortune's Alyson Shontell delves into these issues with Altman in an exclusive interview, offering further insight into the challenges facing the leaders of the most influential AI organizations.