Is ChatGPT safe for business? What Australian organisations should know

Is ChatGPT safe for business? It can be, but "ChatGPT" covers very different products. On OpenAI's individual plans (Free, Go, Plus and Pro), OpenAI may use your conversations to train its models unless you opt out. On ChatGPT Business, Enterprise and Edu and the API, OpenAI says it doesn't train on your inputs or outputs by default, and admins get controls over retention and access. What makes ChatGPT safe or unsafe for your organisation is mostly which plan staff use and what they put into it.

For Australian organisations there's a second question: where the data goes. As of October 2026, OpenAI offers storage at rest in Australia for eligible new ChatGPT Enterprise and Edu workspaces and eligible API customers. It doesn't offer Australian inference residency for ChatGPT, so prompts are processed outside Australia.

ChatGPT plans compared#

Based on OpenAI's help centre, enterprise privacy page and API documentation as of October 2026. Check OpenAI's current documentation before you decide.

Training: personal accounts vs business plans#

OpenAI's help centre is clear about the split. For "services for individuals, such as ChatGPT and Codex", OpenAI says it "may use your content to train our models". Users can opt out by turning off Improve the model for everyone under Settings > Data controls, or through OpenAI's privacy portal. Two caveats apply even after opting out:

• If a user gives feedback, such as a thumbs up or down, OpenAI says the entire conversation linked to that feedback may be used for training.
• The setting belongs to the individual. Your organisation can't see or enforce it on personal accounts.

Temporary Chat isn't a fix either. OpenAI says temporary chats aren't used to improve its models "while they remain temporary". They are still sent to and processed by OpenAI, and saving one turns it into a regular chat.

For business products the default flips. OpenAI states that "by default, we don't use inputs or outputs from ChatGPT Business, ChatGPT Enterprise, ChatGPT Edu, or our API" to improve its models. Its enterprise privacy page adds that customers own their inputs and outputs (where allowed by law). Business data is used for training only if the customer explicitly opts in.

That's a meaningful protection, and it's the main reason a business plan is safer than staff using personal accounts.

Who can see your conversations#

Not training on data is different from nobody seeing it. OpenAI's enterprise privacy page says:

• ChatGPT Business: workspace admins can view, export and delete members' conversations. OpenAI's access is limited to authorised employees for engineering support, investigating potential abuse and legal compliance, and to specialised third-party contractors bound by confidentiality who review for abuse and misuse. Deleted or unsaved conversations are removed within 30 days, unless longer retention is legally required or reasonably necessary to protect the service or others.
• ChatGPT Enterprise and Edu: admins control retention, and authorised OpenAI employees access conversations only to resolve incidents, recover conversations with your permission, or where required by law.
• All business data may be run through automated content classifiers and safety tools.

OpenAI also reports SOC 2 Type 2 audits for ChatGPT Business, Enterprise and the API platform, encrypts data at rest (AES-256) and in transit (TLS 1.2 or higher), and offers a Data Processing Addendum for ChatGPT Business, Enterprise and the API.

Where ChatGPT data is stored and processed#

OpenAI's privacy policy for users outside Europe and the US says it processes and stores personal data on servers "in the United States, or in countries or territories where our affiliates and partners or our vendors and service providers are located".

Data residency changes that only for some customers:

• Storage at rest in Australia is available to eligible API customers and new ChatGPT Enterprise and Edu workspaces, at no extra cost for Enterprise and Education plans. It covers conversations, files, custom GPTs, memory and some other content. Workspace metadata, billing information, user logins, external integrations and some transient processing can still sit outside the region.
• Inference residency, which keeps model processing in-region, is available for ChatGPT only in Europe, the United States and the UAE as of October 2026. OpenAI says that even with it on, CPU processing such as extracting text from uploaded PDF or Word files can still happen elsewhere.
• The API lists Australia for regional storage but not regional processing. OpenAI's documentation notes that where a region doesn't support regional processing, it may process and temporarily store customer content outside the region.

Why it matters: under APP 1, your privacy policy must say whether personal information is likely to be disclosed to overseas recipients, and APP 8 sets rules for cross-border disclosures. The OAIC's AI guidance draws a useful distinction. If you use an AI system with protections so that information entered isn't disclosed outside your organisation, such as to the system developer, that is a "use" rather than a "disclosure" of personal information. Contract terms and settings matter as much as the brand name.

What the OAIC recommends#

The OAIC's guidance on commercially available AI products (October 2024) is the closest thing to an official Australian position. Its key points:

• Privacy obligations apply to any personal information entered into an AI system and to personal information it generates or infers.
• As best practice, organisations should not enter personal information, and particularly sensitive information, into publicly available generative AI tools, "due to the significant and complex privacy risks involved".
• Due diligence on AI products "should not amount to a 'set and forget' approach". Check the terms, whether the vendor can use your inputs, and whether risky features can be switched off.
• Organisations remain responsible for accuracy under APP 10, so staff should check AI output before relying on it.

The OAIC's own example involves insurance staff who paste a customer's claim, including health information, into a public chatbot and ask it to assess the claim. That is exactly the use to rule out.

Is ChatGPT safe for confidential information?#

Not on a personal account. On a business plan, the risk is lower but not zero. Business plans remove training by default and give you admin control, but your information is still processed by a US company, mostly outside Australia, under its terms.

Apply a simple test. Ask whether your client, patient or customer would be comfortable seeing their information pasted into a third-party service in another country. Then ask whether your contract with them, your professional rules or your privacy policy allow it. If the answer to either is no, keep that information out, or use a tool whose terms and data location match your obligations.

Rules to give staff#

1. Use only the approved workspace. No personal ChatGPT accounts for work, including on phones.
2. Know what never goes in: health information, tax file numbers, bank details, passwords and access keys, information about children, and anything covered by legal privilege or a confidentiality agreement, unless your policy expressly allows it for an approved tool.
3. Remove identifiers when you can. Replace names and other identifying details with roles or initials when they don't matter to the task.
4. Check every output. AI can be confidently wrong. The person using it is responsible for what gets sent, filed or relied on.
5. Don't let AI decide things about people. Using AI to score, rank or approve people can trigger new privacy policy obligations from 10 December 2026 and needs human review.
6. Don't connect apps without approval. OpenAI notes that external integrations operate under the third party's own terms and may sit outside data residency.
7. Report mistakes quickly. If someone pastes something they shouldn't have, they should tell the privacy officer the same day.

When an enterprise plan or an alternative makes sense#

ChatGPT Enterprise is a serious product with strong controls, and for many organisations it is a sound choice. Its Australian residency covers storage at rest, not processing. If you need both, look at providers that process in Australia. AusGPT is one option. It runs Anthropic's Claude through Amazon Bedrock's Australian cross-region inference, so requests are processed only in AWS Sydney and Melbourne. Conversations and documents are stored in Australia, and customer data isn't used to train AI models. Read how that works on our Claude in Australia page.

The bottom line#

ChatGPT isn't inherently unsafe, and personal accounts aren't a business tool. A business plan, a clear policy and a few firm rules on information handling make it reasonably safe for everyday work. For sensitive information, decide where it may be processed first, then choose the tool. For the wider legal picture, see our guide to the 2026 Privacy Act reforms.