The piece argues that power in AI is increasingly zero-sum, and leaders in the field could wield influence that outpaces the broader public, prompting calls to curb AI power to prevent an imbalance.

The report places current industry context around Anthropic and potential model updates or re-releases, underscoring openness and governance amid rapid developments.

Open Frontier, a one-day San Francisco gathering hosted by the Laude Institute, brought together AI researchers, CEOs, and academics to discuss the field’s trajectory.

Examples cited—OpenAI’s defense partnerships, a Pentagon contract, Anthropic’s stance on autonomous weapons, and AI firms’ involvement in U.S. political contests—illustrate the widening political influence and lobbying capacity of AI players.

Andy Konwinski framed open AI research as essential for the field’s health and for democracy, warning that access to frontier developments is narrowing.

The event’s central theme was that the AI frontier is closing as access and visibility shrink, driven by decisive moves from leading organizations rather than natural progression.

AI companies are pouring hundreds of billions into pursuing artificial general intelligence, signaling ambitions that could shift power dynamics even without a monopoly.

Corporate power is being framed as potentially more impactful than profits, with concerns about a power-hungry, possibly autonomous AI influencing world events and policy.

There is broad concern about the concentration of AI power in a handful of closed laboratories that control talent, compute, and data.

Konwinski urged attendees to see themselves as active shapers of the landscape, not passive observers, and to push for more open AI research.

The author argues leaders from OpenAI, Anthropic, and DeepMind are garnering power comparable to nation-states, potentially equaling political power at high-level forums.

The conclusion warns that inviting AI leaders to global discussions could reverse if power concentrates further, and emphasizes the need to limit AI power to avert dangerous outcomes.