ChatGPT terms disallow its use in providing legal and medical advice to others (www.ctvnews.ca)

OpenAI has announced that ChatGPT will stop providing legal and medical advice, shifting from offering prescriptive guidance in high-stakes domains to either refusing such requests or delivering only general informational responses with explicit disclaimers. The move reflects growing concern about liability, regulatory scrutiny, and the known tendency of large language models to hallucinate, which can produce incorrect or harmful recommendations in contexts where errors carry real-world consequences. For the AI/ML community this is significant because it highlights how product safety layers, policy rules, and model behavior controls are now driving feature choices as much as capability improvements. Technically, the change will be enforced through updated instruction prompts, post‑processing safety classifiers and refusal-generation mechanisms, and possibly revised fine-tuning/objective functions to reduce risky outputs. The decision accelerates demand for specialized, verifiable domain models, retrieval-augmented pipelines with provenance and human-in-the-loop verification, and stronger calibration/uncertainty estimation. It also sets a commercial and regulatory precedent: general-purpose LLMs may need clearer usage boundaries, while startups and enterprises building domain-specific assistants will be expected to incorporate expert oversight and traceable source attribution.