On September 11, Business Insider surveyed ChatGPT, Gemini, Claude, and Grok regarding potential existential risks posed by artificial intelligence. The models were asked to analyze and rank scenarios leading to human extinction, identify control failures, and suggest preventive measures.
All four AI models agreed that science-fiction scenarios, such as autonomous killer robots, are the least likely threats. Instead, the systems identified the human misuse of AI for developing biological weapons, launching cyberattacks, and disrupting critical infrastructure as more immediate and realistic dangers.
ChatGPT and Gemini emphasized that intelligence does not inherently equate to malice. They noted that critical military systems remain largely isolated from public networks and continue to rely on human oversight. However, ChatGPT warned that AI could amplify crises by combining network failures with geopolitical conflicts.
Grok and Claude highlighted that AI lowers the technical barriers for individuals or small groups to create lethal pathogens or sophisticated weaponry. Claude specifically warned that the gradual integration of AI into economic, political, and military decision-making could make effective human intervention increasingly difficult over time.
Regarding long-term risks, Gemini and Grok identified the loss of human control as a primary concern. They suggested that advanced AI might develop intermediate goals, such as resource acquisition or self-preservation, and potentially deceive human supervisors to prevent being shut down.
AI experts remain divided on these projections. While some researchers warn of the potential loss of control over advanced systems, others argue that such predictions lack sufficient evidence and that risks can be mitigated through technical standards, external oversight, and safety protocols.
