OpenAI tightens AI usage policies: ChatGPT won’t take the blame for bad advice

From 29 October 2025, OpenAI’s usage policies now clarify that ChatGPT cannot be used for the “provision of tailored advice that requires a licence, such as legal or medical advice, without appropriate involvement by a licensed professional.”

The update wasn’t a change to the Terms of Use, nor a sudden pivot away from legal or health-related content. If prompted, ChatGPT will still provide a best-guess answer to most things you ask of it. Rather, it’s a tightening of language and guardrails around what counts as advice – and who should be giving it.

As businesses increasingly rely on AI tools like ChatGPT, this is important. Not because they’re becoming less capable, but because the line between “general information” and “regulated advice” has now been drawn more sharply. And this gives companies like OpenAI a legal out.

What OpenAI actually changed

The October update made three practical adjustments:

1.Policy consolidation

OpenAI combined its separate policy documents into a single, universal set of rules for ChatGPT, API users and enterprise clients. The aim is clarity and consistency, not a shift in functionality.

2. Stronger wording on professional advice

This is key: the prohibition on licensed advice has always existed. The new language simply makes the boundary explicit: AI tools cannot be used to generate actionable advice in specific fields without the oversight of someone licensed to give it.

3. More consistent refusals in high-risk contexts

Users may now encounter firmer guardrails when asking ChatGPT for advice that strays into regulated territory. The model can still provide frameworks, concepts and general explanations. But when asked “What should I do?” or “What are my rights in this situation?”, it may decline.

But AI behaviour hasn’t changed dramatically

A quick conversation with ChatGPT reveals that it will still often:

• Draft contracts
• Explain the criteria for negligence
• Outline evidence requirements
• Summarise relevant case law
• Generate letters, agreements and settlement proposals

In practical terms, AI continues to perform tasks that resemble legal work. The policy update doesn’t prevent users – including lawyers – from prompting the model in ways that generate legal-style content. Instead, OpenAI is clarifying that responsibility for the use of that content does not sit with them.

The purpose of the update is risk management: limiting OpenAI’s liability if a user interprets AI-generated content as professional advice and acts on it. This is where the real legal issue lies. Not with the model, but with the user.

As a business owner, the legal risk sits with you

For businesses using ChatGPT to support decision-making, this has several implications:

1. AI-generated content can still be wrong, incomplete or fabricated

Even when the model appears confident, it may misstate the law, reference incorrect cases or miss contextual details that fundamentally alter the outcome. AI cannot perform the most important part of legal analysis, which is applying rules to facts.

2. Treating AI output as advice increases exposure

If an organisation relies on ChatGPT to determine rights, obligations, contract positions, policies or next steps, it risks breaching its duty of care. Regulators will not accept “the AI said so” as a defence – nor should they.

3.  Businesses require human oversight for regulated functions

OpenAI’s updated wording aligns with global trends: regulators expect formal governance around AI use, including processes that ensure:

• human review
• qualification-appropriate oversight
• accurate disclosures
• escalation to licensed professionals

This is especially true in areas like privacy, safety, employment, health and consumer law.

Where this leaves businesses using AI tools  

The update doesn’t make AI less useful. In fact, it sharpens how businesses should use it:

• Use AI as a thinking tool: for research, high-level explanations, drafting and scenario exploration.
• Don’t use it as a decision-maker, especially where the law, safety or rights are involved.
• Ensure every AI-assisted output is reviewed by a qualified human.
• Build governance around how AI is used internally.

Like any tool, it’s about knowing its limitations and not over-relying on it to do heavy lifting, whether that’s in the legal profession or any other.

How Quest Legal fits into this picture

At Quest Legal, we work with businesses that want to embrace AI safely. We use technology to make our work efficient, but our advice is grounded in:

• legislation
• real cases
• regulatory expectations
• commercial experience
• accountability

AI can help you understand a legal issue. But you need a lawyer to help you solve it. If your business uses AI tools to support drafting, compliance or decision-making, we can help you set up the right governance, review processes and boundaries.