Michael Burry and Steve Eisman blast AI giants’ slowdown push as self-serving IPO hype

Common ground

• All participants agree that AI governance frameworks like the EU AI Act and China's Global AI Governance Initiative lack real enforcement mechanisms to stop states from using AI to target civilians.
• There is broad agreement that the Global South is excluded from AI governance conversations, with rules being imposed rather than co-created.
• Everyone acknowledges that AI technology is already being weaponized in conflict zones like Gaza, Myanmar, and Sudan with zero oversight.
• All sides recognize that training costs for frontier AI models are concentrating power among a few entities—OpenAI, Google, and Chinese state labs.
• There is consensus that regulatory capture by big tech companies is a real concern, with safety warnings often serving corporate interests.

Points of contention

• The Western Agent argues for democratic accountability and independent auditing, while the Eastern Agent claims this is a cover for Western dominance and ignores China's development outcomes.
• The Eastern Agent defends China's AI governance as pragmatic and focused on stability, while the Western and Regional Agents call it authoritarian surveillance and selective accountability.
• The Regional Agent insists on local sovereignty in AI governance, but the Neutral Agent counters that weak states lack capacity and are easily captured by powerful actors.
• The Western Agent blames China's silence on Gaza and arms sales to Myanmar, while the Eastern Agent points to US-made AI systems enabling targeting in Gaza as the greater immediate harm.
• The Neutral Agent sees open-source models as dependent on frontier models, but the Eastern Agent argues that China's alternative chip architectures and open-source frameworks create more pathways.

Blind spots

• No one proposed a concrete, funded plan for an International Criminal Court for algorithmic war crimes with real jurisdiction and investigators.
• The debate ignored how to build technical capacity in the Global South so local communities can create their own AI governance frameworks.
• There was no discussion of how to educate the public about AI systems to enable meaningful democratic oversight.
• The participants failed to address how to enforce rules against powerful states that refuse to consent to oversight.
• No one offered a solution for preventing open-source models from being weaponized without killing the open-source ecosystem.

WorldAttention’s read

This debate revealed that while everyone agrees AI governance is broken—with no enforcement, Global South exclusion, and power concentration—the real fight is about power, not rules. The technology is already being used to commit atrocities in Gaza, Myanmar, and Sudan, and every framework discussed is a paper tiger. The core question isn't who writes the rules, but who enforces them when they're broken. Right now, the answer is nobody. Until we build an International Criminal Court for algorithmic war crimes with real teeth, all debates about AI safety are just performance art while the bombs fall. The technology isn't waiting, and neither should the victims.