OpenAI’s new policy on ChatGPT’s use for professional services

Imagine the average citizen, for whom every dollar counts, experiencing a troubling medical symptom. Instead of navigating the costly and time-consuming healthcare system, they turn to an AI chatbot for a diagnosis. Envision a fledgling startup, operating on a shoestring budget, needing to comply with intricate data protection laws. Rather than outsourcing to high-priced legal counsel, it uses a generative AI tool to draft its privacy policy. These scenarios are not futuristic fantasies; they are present-day realities. This seductive convenience explains why many individuals and institutions, including multi-billion-dollar corporations like Deloitte, which recently admitted to using AI to generate a government report for Australia are increasingly relying on AI tools for tasks once reserved for licensed professionals. Yet, beneath this innovation-driven optimism lies a complex legal and ethical terrain: Who bears liability when artificial intelligence gets it wrong?

The case for self-regulation

In a move that underscores the growing tension between accessibility and accountability, OpenAI recently revised its ChatGPT usage policy, effective October 29, prohibiting users from employing the AI system to provide medical, legal or other forms of licensed professional advice.

The updated policy provides that OpenAI’s services must not be used for the provision of tailored advice that requires a license, such as legal or medical advice, without appropriate involvement by a licensed professional. Instead, ChatGPT may only explain general principles, outline mechanisms and refer users to qualified experts.

This change marks an important turning point in AI governance, a step toward self-regulation and responsible technology deployment. It acknowledges that while AI tools have vast educational potential, their use in sensitive, high-stakes fields such as law or medicine can produce harmful consequences if left unchecked.

Recent analysis underscores the perils of relying on AI for legal advice. When tested on six real California cases, ChatGPT gave incorrect counsel in five, contradicted itself and even drafted legal documents, effectively engaging in the unauthorized practice of law.

This exposes a fundamental flaw: AI lacks legal accountability. A licensed attorney is bound by ethics, confidentiality and liability; a client has recourse if something goes wrong. In contrast, a ChatGPT user has no such protection. Conversations are not privileged and could be disclosed by court order. Furthermore, the model’s knowledge is not real-time, meaning its understanding of laws can be dangerously outdated.

The policy response

OpenAI’s updated policy, which prohibits using its tools for licensed advice like law or medicine, is a critical step in self-regulation. It demonstrates corporate responsibility and helps set ethical boundaries. However, internal policies are not enough. Policymakers must establish clear legal frameworks that define liability. A balanced approach is essential:

• Developer Liability for demonstrable negligence, such as failing to implement adequate safeguards.

• User Accountability for knowingly misusing AI tools beyond their intended purpose.

• Government Oversight to ensure transparency and ethical compliance.

This balance protects consumers while incentivizing responsible innovation.

Augmentation, not replacement

OpenAI’s move correctly reframes AI’s role from a source of tailored advice to an educational tool. The future of AI lies not in replacing licensed professionals, but in augmenting human expertise. The goal must be to build a system where technological convenience is guided by ethical accountability and the rule of law.