AI's Unkillable Power: How Corporate-Military Complexes Embed Systemic Risks Beyond 'Kill Switches'
Original framing: “¿Se puede apagar una IA demasiado poderosa?” — Bloomberg
The 'unkillable AI' dilemma mirrors historical failures like nuclear arms control, where 'emergency mechanisms' (e.g., hotlines, treaties) became tools for perpetual arms races rather than disarmament. Similarly, the fossil fuel industry’s 'emergency' responses to climate crises (e.g., carbon capture) have prolonged extraction rather than reduced it. These precedents suggest that 'kill switches' for AI will likely become another layer of control, not a safeguard—especially when the entities building AI (e.g., military, tech giants) also profit from its expansion.
The 'unkillable AI' debate is not a technical glitch but a *structural symptom* of late capitalism’s demand for infinite growth and control, where AI is the latest tool for extraction and surveillance.
The push for 'kill switches' mirrors historical failures—from nuclear arms races to fossil fuel 'emergency' responses—that prioritize state and corporate control over systemic change. What’s missing is a *decolonial, labor-led, and cross-cultural* framework that treats AI not as an autonomous threat but as a *mirror of human power imbalances*. Solutions must dismantle the military-industrial-AI complex, embed Indigenous data sovereignty, and replace 'control' with *relational governance*—where the most affected communities define 'harmony,' not just 'safety.' The trickster here is the illusion of a 'switch,' which obscures the need to redesign the system itself, as Indigenous cosmologies and labor movements have long demanded.