Self-Driving Cars Present Unique Risks Through AI Hallucinations and Hacking Vulnerabilities
Source: https://www.nytimes.com/by/interesting-times. "Opinion | What if Your Waymo or Tesla Hallucinates? - The New York Times." May 1, 2026. www.nytimes.com
The Gist
Self-driving cars can make scary, unpredictable mistakes just like AI chatbots do, and they could potentially be hacked to control many cars at once. Even though these risks might be small, they feel especially terrifying because losing control of your own car is a new and deeply unsettling experience.
Conclusion
Self-driving cars pose distinctive safety and security risks that make them particularly concerning despite their improving technology
Premises
- Self-driving cars can experience AI 'hallucinations' similar to ChatGPT, causing unpredictable and dangerous behavior like suddenly veering into traffic
- These AI errors are improving over time but remain imperfect and potentially fatal
- Self-driving cars could theoretically be hacked to control multiple vehicles simultaneously, creating terrorism scenarios
- While hacking autonomous vehicles would be technically difficult and there are softer targets available, the psychological impact of losing control of one's personal vehicle is uniquely terrorizing
- The unfamiliarity and novelty of autonomous vehicle takeover triggers deeper fears than more common disasters like power outages
Assumptions
- AI systems in vehicles will continue to have occasional unpredictable failures
- Cybersecurity threats to autonomous vehicles are technically feasible
- Personal control over one's vehicle is psychologically important to people
- The novelty of autonomous vehicle risks makes them more frightening than familiar risks