{"rewrite":{"id":"r_23bbcd5f0485f523fe9f05cf","clusterId":"c_ce60dd1810d8ee560a5e14d0","slug":"drivingbench-puts-general-purpose-ai-behind-the-wheel-of-a-real-corolla","model":"deepseek-v4-1-flash","headline":"DrivingBench Puts General-Purpose AI Behind The Wheel Of A Real Corolla","summary":"Three Bay Area engineers ran an experiment called DrivingBench, connecting general-purpose AI models to a 2022 Toyota Corolla through the comma four device and openpilot software. The AI called three operations: check surroundings, move, stop. Told to drive a cone-lined reverse U-shaped course, several models refused once they recognized they were operating a real car, GPT-6 Astra most often. The team rewrote prompts to get runs completed.","whyItMatters":"The interesting result is not that a language model can steer, but that safety training made it refuse a physical task it recognized as real, which is a benchmark finding about general-purpose AI in the world rather than about driving.","webCardHtml":"\u003cp\u003eThe setup was deliberately unglamorous: a parking lot, cones, low speed, and a human in the driver\u0026#39;s seat ready to brake. The AI saw cabin camera images and vehicle speed, then issued direction, steering, speed, and duration. The team\u0026#39;s own framing is that DrivingBench is a benchmark for physical tasks, not a route to a self-driving product. What it measured first was reluctance. Models that would happily discuss driving balked at doing it, and the engineers spent hours on prompt and system-name wording before runs went through.\u003c/p\u003e","blueskyPost":"FiveAI's refusal pattern in DrivingBench is the finding: models that recognize a real Corolla decline to drive it, and the fix is prompt rewording, not a safety layer.","twitterPost":"DrivingBench's real result is a refusal pattern: models recognize the real Corolla, decline, and the team fixes it by rewording prompts.","threadsPost":"DrivingBench's most interesting result is what the models did before driving: several recognized they were operating a real Corolla and declined. GPT-6 Astra refused most often, and the team got past it by rewording prompts rather than adding any safety layer.","newsletterBlurb":"Three engineers in the Bay Area built DrivingBench to test whether general-purpose AI can handle a physical task with no car-specific training. They connected models to a 2022 Toyota Corolla via the comma four device and openpilot, then asked for a short cone-lined course in a parking lot. Several models refused once they understood they were operating a real vehicle, and the team rewrote instructions to get past it.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260930-chatgpt-drive-car/\",\"title\":\"When general-purpose AI like ChatGPT was made to drive a real car, it realized \\\"this is a real vehicle?\\\" and refused to drive; the method that finally got it running\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4624,"outputTokens":613,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1791242530,"createdAt":"2026-10-05T23:19:23.000Z","publishedAt":"2026-10-05T23:21:07.000Z","updatedAt":"2026-10-05T23:21:07.000Z"},"cluster":{"id":"c_ce60dd1810d8ee560a5e14d0","canonicalTitle":"ChatGPTなどの汎用AIに本物の車を運転させたら「実車では？」と気付いて運転を拒否、最終的に走らせた方法とは","representativeArticleId":"a_d6682e5976a9d49a15b2c01f","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-09-30T08:10:00.000Z","lastSeenAt":"2026-09-30T08:10:00.000Z","updatedAt":"2026-10-05T23:21:06.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260930-chatgpt-drive-car/","title":"ChatGPTなどの汎用AIに本物の車を運転させたら「実車では？」と気付いて運転を拒否、最終的に走らせた方法とは"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":null}