A team has built AI-Sinkhole, software that uses machine learning to automatically detect and block access to chatbot services on school networks during proctored exams. Schools deploy it like a network firewall that learns what LLM services look like and cuts them off when tests are running.
Why it matters
This is a narrow technical response to a real classroom problem: students using ChatGPT and similar tools to bypass exam integrity. The interesting signal is not the tool itself, but what it reveals about how institutions respond to AI — they build more AI to police AI, rather than changing assessment methods or exam design. Schools are automating detection and enforcement instead of redesigning what tests measure.
The signal
Whether schools actually deploy this at scale, or whether the blocking becomes an arms race (students find new services, the blocker learns them, repeat).