Generative AI tools like OpenAI (aka ChatGPT) have revolutionised how we work, making it easier than ever to generate insights and process information. But when it comes to business-critical data analysis, the convenience of public AI tools comes with significant risks.
If your organisation handles sensitive, high-value, or regulated data, as is common in the liquor industry, relying on a public AI model could leave you exposed.
What do you need to consider?
Data Security & Privacy Risks
When you input data into a public AI model, you often lose control over where it’s stored and how it’s processed.
- Risk: Sensitive or commercially valuable datasets may be transmitted and stored on external servers, and potentially used to train future models.
- Impact: This could breach privacy laws, contractual obligations, or internal compliance policies.
The Scale of the Risk: What the Data Shows
The concern about public AI tools and business data is not theoretical. Industry research consistently highlights the exposure:
| Fact / Risk | Detail |
|---|---|
| 68% of employees who use AI tools at work | have entered sensitive business information into public AI systems (Cyberhaven, 2024 AI Data Leakage Report) |
| OpenAI’s default data retention policy | allows user inputs to be reviewed by staff and used to improve models unless enterprise opt-out settings are explicitly configured — a step most SME users never take |
| The Australian Privacy Act (Privacy Act 1988) | requires businesses to take reasonable steps to protect personal information. Sending customer or employee data to an overseas AI platform without a Data Processing Agreement may constitute a breach |
| The liquor industry’s data | including scan-level sales data, supplier rebate structures, and wholesale pricing — is commercially sensitive intellectual property. Once submitted to a public model, there is no contractual guarantee it will not influence future model outputs accessible to competitors |
Public AI vs. Purpose-Built Industry Data Tools: A Practical Comparison
← Scroll to see full table →
| Factor | Public AI (e.g., ChatGPT) | OnTap Data |
|---|---|---|
| Data storage | External servers; may be used for training | Hosted and controlled by OnTap; never used for third-party training |
| Industry specificity | Generalist; no liquor industry data structure | Built specifically for liquor suppliers, wholesalers, and buying groups |
| Audit trail | No reproducible audit trail | Full audit trail for compliance and reporting |
| Integration | Manual copy-paste; no live data connection | API integration with Impos, H&L, Bepoz, Zoho, Salesforce, and more |
| Accuracy | Prone to hallucination on numerical data | Validated outputs based on structured, verified industry data |
| Privacy Act compliance | Dependent on user configuration | Designed for Australian regulatory requirements |
Accuracy & Reliability
Public AI models are designed to predict language, not to guarantee analytical accuracy. They can present incorrect or fabricated numbers with complete confidence, a phenomenon known as “hallucination.”
- Risk: Inaccurate data analysis can lead to flawed business decisions.
- Impact: Lost revenue, poor stock management, or reputational harm.
Loss of Proprietary Advantage
Your business data is intellectual property. Sharing it with public AI tools risks diluting its competitive value.
- Risk: Unique market insights could indirectly benefit competitors if data influences broader model training.
- Impact: Loss of market edge and reduced control over proprietary knowledge.
Governance & Auditability
With public AI tools, outputs can vary between identical queries, making it hard to track or replicate analytical decisions.
- Risk: No clear audit trail for compliance reviews or internal investigations.
- Impact: Difficult to justify decisions to regulators, stakeholders, or auditors.
Generalist Analysis
Public AI is generalist by nature. While it can answer almost anything, it doesn’t inherently understand specific industries’ data structures, supplier nuances, or POS system integration.
OpenAI can be useful for brainstorming or summarising non-sensitive data, but when accuracy, security, and compliance matter, a custom enterprise solution is the smarter investment.
With OnTap Data you get:
- Full data security and compliance control.
- No third-party training use, your data remains confidential and is never used to improve models for other users.
- Accurate, verified outputs based on industry-specific data with a structured, verified database tailored to your industry
- Built-in validation rules (including that all-important human touch) to ensure output accuracy before it reaches decision-makers.
- Tailored functionality that matches the exact needs of your business’s operations.
- Seamless integration with POS systems like Impos, Bepoz, H&L, Shopfront, Blueize, SwiftPOS and CRM systems including Hubspot, Zoho, Salesforce and ForteIS
- Constant updates to match changes in product catalogues, supplier information, and regulatory requirements.
Confidence, reliability and accuracy are at the core of everything we do. Don’t risk your critical business data to open-source platforms; talk to us about a tailored, secure data solution that scales with your needs.
FAQ
Does OpenAI use my business data to train its models?
- By default, OpenAI may use inputs from non-enterprise accounts to improve its models. Enterprise accounts can opt out, but this requires a paid plan and explicit configuration. Many businesses using the free or standard ChatGPT interface are unaware their data may be retained and reviewed.
Is it legal to send customer data to ChatGPT in Australia?
- It depends on the nature of the data and whether you have a Data Processing Agreement in place. Under the Australian Privacy Act, businesses must take reasonable steps to protect personal information. Sending identifiable customer data to an overseas platform without appropriate safeguards may constitute a breach. Always seek legal advice for your specific situation.
What is a safer alternative to using public AI for liquor industry data analysis?
- Purpose-built platforms like OnTap Data process your sales, supplier, and promotional data within a secure, industry-specific environment. Your data is never used to train external models, outputs are validated against structured databases, and the system integrates directly with your existing POS and CRM platforms.
