Insight,

Future ready: deadline approaches for Government to be AI trained

AU | EN
Current site :    AU   |   EN
Australia
Singapore

The deadline for Commonwealth entities to train their people in Artificial intelligence (AI), including generative AI, is fast approaching. With the Australian Government’s commitment to leveraging AI to improve public services comes the need to ensure the APS understands the technology, its benefits and its risks. In this article we provide an overview of the AI training requirements set by the Digital Transformation Agency (DTA) and provide our top tips on developing and implementing that training.

This is the second in our series of articles on AI in Government. In case you missed our first article, you can find out about the requirements for AI Transparency Statements here. A reminder those Transparency Statements are mandatory for all non-Corporate Commonwealth entities and they’re due by 28 February 2025.

Key takeaways

  • While the Digital Transformation Agency (DTA) hasn’t gone so far as to make staff training in AI mandatory, they have strongly recommended all Commonwealth entities implement, by 28 February 2025:
    • “fundamentals” AI training for all staff, and
    • additional AI training for staff who have particular roles and responsibilities such as the procurement, development, training and deployment of AI systems.
  • DTA has developed Guidance for staff training on AI (the Guidance) and a fundamentals AI training module which is available through the APS Learning Bank.
  • The fundamentals training is intended for all APS staff and contractors to “help staff become confident using AI in a responsible way”. Agencies can implement this training and/or develop their own training programs, geared at empowering their teams to utilise AI effectively, by understanding the benefits, and being alert to potential risks.
  • The DTA’s fundamentals training doesn’t cover the “additional” training contemplated for those with roles that bring them into contact with procurement, development, training and deployment of AI systems. For that, Agencies will need to decide (quickly) what additional training is needed for those people, and how they will provide it.

Why staff need AI training

AI is already used in many areas of the APS. The Australian Government is committed to further using AI to assist in improving productivity of the delivery of government services.

To ensure the effective and ethical take-up of AI it is important that the APS understands what AI is and how it works (at a high level), including its uses, limitations and biases. For example, staff need to understand what happens when they enter information (or “prompts”) into a public AI system like Chat GPT, and that this is a disclosure of that information, so caution needs to be exercised by staff who have access to public AI systems.  Of course, AI training will need to adapt and evolve as AI adapts and evolves and different uses of AI in the public sector are implemented.

What training has the DTA developed for Agencies to use?

DTA has developed an AI fundamentals training program which is available through the APS Learning Bank. This training is aimed at an approximate 20-30 minute completion time and is focused on generative AI. Generative AI creates new content (think ChatGPT creating a new policy proposal). Traditional AI performs specific tasks intelligently but follows rules (think automated decision making according to legislative rules). The fundamentals training covers:

  • An introduction to AI
  • Explanation of generative AI
  • Foundations of safe and responsible use

The DTA will be updating the fundamentals training as the AI landscape changes, so Agencies should consider running the training regularly to keep pace.

The fundamentals training is high level and introductory. While we think this it is a useful introduction to AI (particularly generative AI) for the public sector, as AI take-up expands in the public sector, it’s inevitable that further training and information will be needed. DTA strongly recommends additional training for staff based on their roles and responsibilities, considering especially those people who will be involved in implementation of AI within Agencies. So, think of fundamentals training as a first step for your Agency’s staff. Remember to record training attendance, so you can let the DTA know about attendance if you’re asked.

Agencies need to consider what additional training they will need to provide to other cohorts, based on their interaction with AI. So, in addition to AI fundamentals training for all staff, Agencies should consider whether the following cohorts need a bit more than the 20-30 minute fundamentals training:

  • Senior Executives, who will oversee use and compliance, decide when the use of AI is appropriate and make procurement decisions
  • IT/Technical staff, who will deploy AI systems and need to be able to assist in system development, training, issue resolution and instructions on the use of implemented AI
  • IT Procurement staff, who will need to know more about AI systems to assist in informed purchasing of them and in vendor management

As far as possible AI training should be practically focused. In the table below, we set out some key considerations for developing AI training.

SO THINK…
INDIVIDUAL
Example uses 2
AI Fundamentals
  • How will your Agency upskill all its people?
  • Will you roll out DTA’s AI fundamentals training, or do you have something specific for your environment?
  • Is broader foundational training required for all (or some) staff based on specific Agency activities? If not now, maybe later when there is more use of AI within the Agency, or prior to roll-out of AI technology?
What are the risks of Generative AI in your Agency’s context?
  • Consider the level of understanding that your people will need about the risks of AI, based on the activities your Agency is likely to use it for.  For example, is there likely to be use of AI in higher-risk scenarios such as AI being used to impact users of your services? In that case, staff may need to understand the importance of human oversight.  
  • Is risk training needed that is focussed on potential use of AI by different cohorts of Agency staff
  • Some of the key risks that might need to be understood are:
    • Performance risks (misleading and inaccurate outputs)
    • Compliance risks (privacy, automated decision making and administrative law, IP, confidentiality)
    • Use risks (reputational risk, user misuse, security and safety)
    • Trust risks (bias, discrimination and lack of transparency)
Data governance and Management
  • What training do your people need on integrating AI into existing frameworks and work methods? What help is available within and outside your Agency?
  • Training to ensure ongoing compliance with the Australian Privacy Principles in relation to collection, use, storage and disclosure of data
  • Whether new record keeping processes have been implemented for AI use. How are staff required to keep records of their use of generative AI systems?
  • Training to ensure your Agency’s data security, integrity and compliance protocols are followed
  • The requirements of the Protective Security Policy Framework, Cyber Security Guidelines and 2023-2030 Australian Cyber Security Strategy as they relate to AI use
AI ethics and accountability
  • Training on ethical AI use, mitigating bias, and transparency (including the approach your Agency is taking to your Transparency Statement)
  • Training focused on ensuring ongoing compliance with existing laws, including administrative law, privacy/data, FOI law, discrimination acts, copyright. What do staff need to know about how AI use impacts these areas?
  • Understanding the implications of AI use on individuals and communities
  • Implementing Australia’s AI Ethics Principles
  • What is an AI Impact Assessment, who should complete one and when? Training on how to conduct AI Impact Assessments, what protocols and tools has the Agency implemented. Who can help staff? 
Technical skills development
  • Specific technical training for key cohorts (for example, advanced user/developer training)
  • Is specific training required for staff involved in developing new Agency protocols and procedures for AI use?
  • Is ‘train the trainer’ training required?
  • Continuous learning as AI technology evolves. Who is responsible for AI training development and how will changing use and needs be monitored? 
Change management and communication
  • Internal change management for AI use within government. Will Agency leaders need support and training in implementing AI and “selling” the change? What are the good examples of this?
  • Improving public trust of AI through transparency and engagement – how do staff assist in ensuring transparency and engagement?
  • Is training required to build skills for communicating to the public and non-technical stakeholders about AI processes and outcomes? 

What should Agencies do now?

Right now, we expect that Agencies would (and should) be:

  • assessing current AI skills across key Agency cohorts
  • implementing the DTA’s fundamentals training program for all staff (and recording attendance)
  • developing Agency training programs for additional and specific needs, including considering who will develop and deliver that training
  • monitoring and evaluating AI use by the Agency to ensure ongoing training needs are met

And remember, the DTA strongly recommends your training programs are implemented by 28 February!