Emergency officials from San Francisco and Austin told NHTSA privately that Waymo vehicles are blocking fire stations, freezing more often, and committing more traffic violations than before.
In a private meeting with NHTSA obtained by WIRED, first responder leaders from San Francisco and Austin reported deteriorating Waymo AV performance — vehicles freezing, blocking fire stations, and committing more traffic violations. SF Emergency Management Director Mary Ellen Carroll cited 'backsliding' on previously improved behaviors. Waymo declined to attend the meeting, leaving two reserved front-row chairs empty. Remote operator response times have reached up to three minutes.
This story exposes a core AV engineering failure: behavioral regression at scale. Waymo's edge-case handling — freeze logic, remote operator handoff latency, gesture recognition — is degrading in live deployment, not improving. For developers building autonomy stacks or AV-adjacent tooling, this confirms that continuous behavioral evaluation pipelines and regression testing in real-world conditions are non-negotiable, not optional.
If you're building any safety-critical AI system with human-in-the-loop fallback, benchmark your handoff latency against Waymo's reported 3-minute remote agent connect time — that's your bar for 'unacceptable' in regulated environments.
Suggested exercise · not independently tested. Access requirements and costs may vary.
Go to claude.ai and open a new conversation
Paste this prompt: 'I'm building an AI system with a human-in-the-loop fallback. Waymo's remote operator handoff reportedly takes up to 3 minutes in emergencies. List the top 5 architectural decisions that would reduce that latency to under 30 seconds, with trade-offs for each.'
Review the architectural options against your current system design
Tags