Executives from OpenAI and Anthropic are conducting private simulations of potential AI-driven disasters. These exercises focus on scenarios where critical infrastructure, including banking, power, water, and internet services, faces disruption.
Industry insiders suggest that a major AI-related event, likely involving a cyberattack, could occur within the next six to 12 months. OpenAI stated that these drills serve as planning tools rather than predictions of inevitable outcomes.
Recent security incidents highlight these risks. CrowdStrike reported that a hacker utilized Chinese AI models to breach South Korean banks and consulted Anthropic’s Claude Code tool regarding the sale of stolen data.
AI companies are preparing for significant political and public backlash following any major harm caused by their technology. Executives are currently briefing lawmakers to influence future regulations that may emerge after a crisis.
Legislative proposals currently under discussion include a potential ban on superintelligence and requirements for development pauses. The proposed AI Kill Switch Act would grant the Department of Homeland Security authority to shut down dangerous systems.
