Artificial intelligence promises to transform government operations, from fraud detection and case management to public-facing services. But while agencies plan their technology investments, are they giving equal consideration to the people required to secure them?
The workforce challenge is already evident. According to the Australian Public Service Commission’s State of the Service Report 2024-25, 85% of agencies reporting critical skills shortages identified digital and ICT capability as an area of concern. Among those agencies, 79% identified cybersecurity shortages and 53% reported shortages in AI model training and development.
These figures don’t specifically measure AI security expertise, but they do expose a potential problem: Government is expanding its use of technologies that depend on skills that are already in short supply – and AI introduces security challenges that conventional cybersecurity expertise may not fully address.
AI introduces risks that demand different expertise
Traditional cybersecurity controls remain essential, however, AI systems introduce additional vulnerabilities involving model integrity, data provenance and adversarial manipulation.
Prompt injection, for example, can exploit how large language models process instructions and external content. Data poisoning can compromise model behaviour by manipulating training data or feedback. These threats are recognised in the Australian Signals Directorate’s Guidelines for Secure AI System Development.
Consider an agency introducing AI to assist with processing citizen applications. Its infrastructure may meet established security requirements, but can its existing team assess whether malicious inputs could manipulate the system’s outputs or expose sensitive information? Answering those questions requires expertise spanning machine learning, cybersecurity engineering and risk management. The challenge is that these capabilities don’t necessarily exist within the same professional.
An experienced cybersecurity engineer may understand cloud security and threat detection but have limited experience assessing machine-learning models. An AI engineer may understand model architecture without the expertise to identify sophisticated security vulnerabilities.
AI governance adds another dimension, requiring professionals who can translate policy, risk and accountability requirements into practical safeguards. That distinction matters because AI engineering, AI security and AI governance are complementary disciplines, not interchangeable roles.
Why government faces an additional recruitment challenge
Finding professionals with this combination of skills can be difficult. For government agencies operating in classified or sensitive environments, security-clearance requirements can narrow the available candidate pool even further – and traditional recruitment practises may compound the problem.
A project receives funding, a technical requirement emerges and recruitment begins, yet an agency advertising for an ‘AI Security Specialist’ may discover that candidates possess vastly different capabilities. Some specialise in securing the cloud infrastructure supporting AI workloads. Others focus on adversarial testing, application security, model assurance or governance. A poorly defined position description can therefore attract unsuitable candidates or seek an unrealistic combination of expertise.
Meanwhile, project delivery continues and results are still expected.
The solution isn’t necessarily to recruit faster, it’s to identify the security outcomes required before determining which specialists are needed.
This is becoming particularly relevant as Commonwealth AI governance requirements mature.
The Digital Transformation Agency’s updated Policy for the responsible use of AI in government, effective from December 2025, introduces requirements covering AI strategy, accountability, staff training and risk-based impact assessments. These requirements apply to non-corporate Commonwealth entities, subject to specified exceptions.
Meeting governance obligations requires more than documented policies. Agencies also need people capable of implementing controls, assessing technical risks and maintaining effective oversight.
Five workforce questions agencies should be asking
Before committing to their next AI initiative, public-sector leaders should consider five questions.
1. What AI-specific security risks does the proposed system introduce?
A productivity assistant and an AI system supporting high-impact decisions have different risk profiles. Security requirements should reflect the system’s purpose, access to information and potential consequences of failure.
2. Can our existing cybersecurity team assess those risks?
Cloud security, application security and threat-modelling skills provide a valuable foundation, however agencies should identify where AI-specific training or specialist expertise is required.
3. Have we clearly separated engineering, security and governance responsibilities?
Who will develop the system? Who will independently assess its security? Who will own the risks? Clear accountability reduces the likelihood of vulnerabilities being overlooked.
4. Are our clearance requirements proportionate to the work?
Appropriate clearances are essential where required. However, distinguishing between classified and unclassified responsibilities may allow agencies to access a broader range of specialists without compromising security.
5. Are we planning for future capability or just filling immediate vacancies?
As AI becomes more deeply embedded in government operations, requirements for model assurance, adversarial testing, continuous monitoring and incident response may grow. Workforce planning needs to anticipate these demands.
Building capability before it becomes a crisis
Addressing the AI security skills gap doesn’t necessarily require creating entirely new teams. Existing cybersecurity professionals may be able to develop relevant AI security capabilities through targeted training. Specialist contractors can support implementation, independent assessments or short-term projects, while permanent employees retain organisational knowledge and ongoing accountability. For agencies with demanding clearance requirements, developing existing cleared personnel may be particularly valuable.
The most effective approach will depend on the sensitivity, complexity and scale of each agency’s AI systems.
Importantly, external specialists should complement rather than replace internal accountability. Agencies must retain sufficient expertise to evaluate advice, oversee suppliers and understand the risks they accept.
The next 24 months present an opportunity to make these decisions deliberately, rather than under pressure following a security incident or delayed deployment.
Technology procurement, cybersecurity assurance and workforce planning should be considered together, from the earliest stages of an AI initiative, because an AI strategy without an AI security workforce strategy is an incomplete plan.
Sarah Friggieri is a communications specialist and Founder of Sierra Foxtrot Empire. She also works with Needus, an Australian recruitment and consulting business specialising in AI, cybersecurity and technology talent, including security-cleared professionals.
Published by
About our partner
Needus
Needus is a specialist recruitment and executive search partner connecting government organisations with exceptional technology, cybersecurity, digital, data, AI and executive talent across Australia.We are an approved supplier to the NSW Government under SCM0007 Contingent Workforce and SCM0012 Talent Acquisition, enabling Needus to provide NSW Government agencies with both contingent and permanent talent across specialist, technical and executive roles.Our network includes professionals holding NV1, NV2 and PV security clearances, supporting government agencies with roles requiring highly trusted and security-cleared capability.Combining deep industry expertise, proactive headhunting and a highly curated network, we deliver quality candidates quickly and help public-sector organisations build high-performing teams with confidence.
Learn moreHelp your peers
Share what you've learned with fellow public servants