Responsible AI Policy

Purpose, status and scope

Talk Today Consultancy (ABN 85 659 369 847) provides AI, digital, RTO, training and business consulting services in Australia. This policy describes the principles and safeguards we apply when we select, configure, develop, recommend or use an AI system in our operations or services.

It applies to our personnel and contractors, our website tools and AI Readiness Check, and AI-related work undertaken for clients. Project contracts, scopes, security requirements and client policies may impose additional requirements.

We aim to follow a risk-proportionate approach informed by the Australian Government’s responsible AI guidance. That guidance is voluntary and does not replace applicable law. Depending on the activity, relevant Australian requirements may include privacy, consumer protection, anti-discrimination, copyright, confidentiality, employment, recordkeeping and cyber-security obligations.

Our responsible AI principles

Our approach is built around the following commitments:

• Accountability: a person remains responsible for approving AI uses and responding to material issues;
• Benefit and proportionality: AI should address a legitimate need and its expected benefit should justify its risks and costs;
• Human-centred use: AI should support human capability, dignity, autonomy and wellbeing;
• Fairness: foreseeable risks of unfair bias, exclusion or discrimination should be identified and addressed;
• Privacy and security: information and systems should be protected throughout the AI lifecycle;
• Reliability: tools should be suitable for their intended context and tested in proportion to risk;
• Transparency: significant AI involvement, limitations and assumptions should be communicated where relevant; and
• Contestability: people should have a practical way to raise concerns about AI-supported outcomes that materially affect them.

How Talk Today Consultancy may use AI

Subject to appropriate approval and controls, we may use AI to assist with:

• research, brainstorming, drafting, summarising and administrative productivity;
• workflow analysis, automation design, software development and quality assurance;
• document, training-resource and communication support;
• data analysis and business insights where the data use is authorised;
• client prototypes, demonstrations and configured AI solutions; and
• our rules-based AI Readiness Score and related recommendations.

AI output is treated as assistance—not automatically as fact, professional advice or a final deliverable. The level of review depends on the purpose, foreseeable impact and agreed scope.

AI solutions and advice for clients

For client engagements, we aim to:

• define the intended purpose, users, operating environment and success measures;
• identify material risks, limitations, data flows and required human controls;
• select tools appropriate to the use case and the client’s risk tolerance;
• test and document the solution to the level agreed in the engagement;
• communicate known dependencies and limitations; and
• support training, monitoring or review where included in scope.

Clients remain responsible for their decisions, lawful authority to provide data, internal approvals, user access, ongoing operation and any use outside the agreed scope unless our contract expressly assigns a responsibility differently.

Uses we do not permit

We do not knowingly use or support AI for unlawful, deceptive or harmful purposes. Without appropriate legal authority, safeguards and express approval, our personnel must not use AI to:

• make a final high-impact decision about employment, credit, insurance, healthcare, education access, essential services or legal rights;
• discriminate unlawfully or target a person based on protected or sensitive characteristics;
• impersonate a real person, create deceptive deepfakes, manipulate reviews or misrepresent AI output as verified human work;
• conduct unauthorised surveillance, profiling, biometric identification or scraping;
• generate malicious code, fraud, harassment or content that facilitates harm;
• submit passwords, authentication secrets or information classified as restricted to an unapproved AI service; or
• use content in a way that knowingly infringes privacy, confidentiality or intellectual-property rights.

Risk assessment and human oversight

Before material AI use, we consider the intended purpose, affected people, data sensitivity, reversibility, accuracy needs, security, legal requirements and consequences of failure. Higher-risk uses require stronger controls and may require specialist legal, privacy, security or subject-matter advice.

Appropriate controls may include:

• a named accountable owner and approval process;
• qualified human review before action or publication;
• testing against defined performance and safety criteria;
• restricted access, logging, monitoring and fallback processes; and
• a way to correct, pause or withdraw the system when material issues arise.

Data, privacy and confidentiality

Personal information handled through AI is managed in accordance with our Privacy Policy and applicable Australian privacy obligations. We aim to minimise inputs, use information only for an authorised purpose and consider where providers store, access or reuse data.

We do not intentionally enter sensitive personal information, client-confidential information, learner records, health information, credentials or commercially restricted material into a public or unapproved generative AI tool.

Client information will not knowingly be used to train a public AI model unless this is clearly disclosed, lawfully authorised and expressly agreed. Where a service provider may retain prompts or outputs, use data for improvement, or process information overseas, we assess that position and communicate material implications where relevant.

Fairness, inclusion and accessibility

AI systems may reproduce errors or bias from their data, design or use. Where an AI-supported process may affect people, we consider whether inputs, testing and outcomes are appropriate for the relevant population and whether the process could create unlawful discrimination or unreasonable exclusion.

Where proportionate, we use diverse test scenarios, review patterns of error, provide alternative human channels and seek feedback from affected users. We do not claim that a general-purpose AI tool is bias-free.

Accuracy, reliability and transparency

AI can produce incomplete, outdated, biased or confidently incorrect output. We require review proportionate to the context and do not rely on AI alone where accuracy is critical. Important claims, calculations, citations, code or recommendations should be checked using suitable sources, tests or professional judgement.

We aim to tell clients or users when AI involvement is material to a service or outcome, especially where a reasonable person would expect to know. We communicate known limitations and do not make unsupported claims about an AI system’s capability, compliance, accuracy or likely results.

Security, approved tools and suppliers

We consider cyber-security and supply-chain risks when selecting and operating AI tools. Depending on risk, this may include reviewing provider terms, privacy and data-use settings, hosting location, access controls, model limitations, incident processes and the ability to delete or retrieve information.

We use access controls and secure configurations appropriate to the service, keep software maintained, restrict credentials and monitor material systems where feasible. AI systems remain exposed to risks such as prompt injection, data leakage, malicious content, model changes and service disruption; no system can be guaranteed completely secure or error-free.

Intellectual property and content rights

AI inputs and outputs can raise copyright, trademark, licensing, attribution and confidentiality questions. We aim to use source material lawfully, respect applicable licences and avoid knowingly requesting or reproducing protected material in an infringing manner.

AI output may not be unique and ownership or protectability may be uncertain. Unless a client agreement states otherwise, no assurance is given that AI-assisted output is exclusive, registrable or free from all third-party rights. Material intended for commercial publication should receive an appropriate rights and quality review.

AI Readiness Check and score

Our AI Readiness Check uses the answers you submit to calculate an indicative, rules-based score and general recommendations. It is designed as an educational and preliminary business-planning tool.

The score is not a technical, legal, privacy, cyber-security, financial or regulatory audit. It is not a guarantee of performance or suitability and is not used by us to determine legal rights, employment, credit, insurance, access to essential services or another decision that could significantly affect an individual.

Our team may review submitted answers to understand an enquiry and discuss relevant services. Please see our Privacy Policy for information about how assessment responses are handled.

Questions, concerns and AI incidents

If you believe an AI-supported service has produced a material error, unfair outcome, security issue or misuse, please contact us promptly with the relevant context and the outcome you are seeking. We will assess the issue proportionate to its seriousness and may pause use, preserve relevant records, correct output, notify appropriate parties or seek specialist assistance.

Where an outcome materially affects a person and we control the process, we aim to provide a reasonable avenue for human review. Legal, privacy or cyber-security notifications will be made where required by applicable law.

Governance, review and policy changes

Responsibility for this policy sits with Talk Today Consultancy’s management. We review it periodically and when material changes occur in our services, AI systems, risk profile or Australian requirements. We may update the published version and will show the current effective date.

This public policy is a statement of our approach, not a certification, warranty or substitute for a project-specific assessment. If this policy conflicts with a mandatory legal requirement, the legal requirement applies.