Cal Fire’s Chatbot Fails Users With Dangerously Outdated Wildfire Information

While California residents face growing wildfire threats, Cal Fire’s new chatbot is falling dangerously short of its mission to provide critical emergency information. The digital assistant, meant to keep people safe during emergencies, is consistently delivering wildfire containment data that’s up to six days out of date. Not exactly helpful when flames are rushing toward your neighborhood.

The chatbot’s failures extend beyond stale updates. When users ask about evacuation orders—you know, the life-or-death information people desperately need—the system simply can’t answer. Instead, it redirects worried residents to websites or offers vague uncertainties. Great. Check a website during a power outage. Super helpful.

Testing revealed the tool struggles with basic consistency. Ask about evacuation kits and you might get a detailed list—but only if you use those exact words. Try “go bag” instead and you’ll get nothing useful. The chatbot’s keyword recognition is apparently stuck in 2005.

The chatbot can’t handle synonyms—use “evacuation kit” and you’re safe, say “go bag” and you’re on your own.

Cal Fire job openings? The bot says zero vacancies exist when positions are actually available. Real-time emergency developments? Don’t count on it. The system’s reliance on static data sources means it’s about as current as last week’s newspaper.

Experts point to insufficient pre-launch testing as a major cause of these failures. The system, developed by Citibot, is expected to be maintained through 2027 despite its current shortcomings. Experts have strongly recommended that independent third-party evaluations be conducted to properly assess the benefits and risks of such AI systems. Nobody bothered to check if the chatbot could address actual human needs before deploying it on a public facing actual fires. Shocking.

The implications for public safety are serious. Outdated information could delay evacuations or cause residents to make dangerous decisions based on incorrect data. Trust in government emergency services erodes with each inaccurate response.

The chatbot’s shortcomings highlight a critical gap between technological promises and practical emergency management. California’s fire seasons grow more intense yearly. Residents deserve better than a half-baked digital assistant that can’t tell them whether they need to flee their homes.

Maybe start with the basics next time?