Comment by bcjordan

Not sure about their policy/internals myself, but since GPT-3.5-turbo is 1/10th the cost of the previous model it's really hard now to run up a significant (to a funded startup) bill via legitimate usage — especially since at this phase any legit usage may be worth its weight in lessons about how the tool can be used. May be worth rate limiting / DoS protection but even then, could just cap OpenAI spending and spend the time to patch it if/when it happens (e.g. import @upstash/ratelimit or adding a quick Cloudflare rule)