OpenAI is going Meta route, as it considers memory-based ads on ChatGPT (www.bleepingcomputer.com)

OpenAI is internally considering adding memory‑based ads to ChatGPT as it looks for reliable revenue ahead of a potential public listing. The move follows pressure from weak conversion—ChatGPT reportedly has ~800 million users but only ~5% pay—and heavy spending plans despite an estimated ~$13 billion revenue stream that largely comes from paying customers. The company already offers tiered plans (Go, Plus, Pro, enterprise) and has begun widening its low‑cost Go plan (now €4/$4/£3.50) into several European countries. It also introduced purchasable credits for Codex and Sora at $40 per 1,000 credits. Internal debates and focus‑group findings that users assume sponsorship or ranked answers, plus a large influx of former Meta staff (~630 ex‑Meta employees), reportedly bolster arguments for ad monetization. For the AI/ML community this raises technical and ethical questions: memory‑based advertising implies persistent, personalized context stored and used to target or rank responses, affecting model behavior, data governance, privacy, and transparency. If sponsored content can influence outputs, it could introduce bias, reduce trust in model neutrality, and complicate evaluation/benchmarking. Practically, monetization via ads may steer product design toward consumer features and segmentation of model access (e.g., Go lacks access to premium “Thinking” models), shifting resource allocation away from pure AGI research toward scalable, privacy‑safe personalization and content‑ranking systems.