3.2.
2 Ensure Accountability and transparency
ACTION ACCOUNTABLE RESPONSIBLE CONSULTED INFORMED
Clearly define and communicate
responsible AI-related authorities
within the organisation, including
Management
governance, assurance, procurement, Executive level
Level
- Everyone
ethics, cyber, privacy, legal,
technology, data governance, and risk
management.
Review and endorse the AI Assessment
Framework (AIAF) compliance plans [1],
Management
detailing Department/Agency progress Executive level
Level
- -
towards compliance in ensuring use of
the AIAF.
Ensure each AI solution has
documented accountabilities for
Management Product Executive
managing risks, ensuring continuity, Level
Product Owners
Owners level
enabling appeals, and providing
evidence for decisions and actions.
Ensure record-keeping for decisions
related to managing the risk of AI Management
Executive level Product Owners -
solutions such the results of applying Level
the AIAF and risk mitigations.
Publish regular transparency reports
for high-risk or customer-facing AI
Management Product
systems, detailing use cases, Executive level
Level Owners
Everyone
performance, governance practices,
and any incidents or interventions.
[1] Compliance plans communicate departmental compliance with the AIAF, tracking implementation
progress and raising awareness of challenges for support. Reporting requirements apply to
department Governance and Assurance functions and are collated by Digital NSW through the AI
Secretariat.
Understanding Responsibilities in AI Practices 4
Allocate Resources
ACTION ACCOUNTABLE RESPONSIBLE CONSULTED INFORMED
Support initiatives to increase AI risk
management awareness and Management
Executive level Everyone Users
capabilities at all levels of the Level
organisation.
Allocate budget and resources for
responsible AI, including expert Management Product
Executive level Executive level
advisory services (legal, data, privacy, Level Owners
ethics, technology, risk).
Provide sufficient training and tools for Management
Product Owners - Everyone
ethical AI implementation. Level
Ensure adequate resources for
Management Product
continuous monitoring and evaluation Executive level
Level Owners
-
of AI systems that could cause harm.
Reduce costs of digital governance
and assurance with streamlined,
integrated processes across Management Product
Executive level Everyone
cybersecurity, privacy, ethics, legal, Level Owners
AI, data governance, and technology &
architecture domains.
Understanding Responsibilities in AI Practices 5
3.2.3 Ensure Compliance and Risk Management
ACTION ACCOUNTABLE RESPONSIBLE CONSULTED INFORMED
Ensure governance and assurance
oversight for compliance with AI-
related laws and regulations (e.g.,
human rights, privacy, data Management
Executive level Product Owners -
protection, administrative law, Level
consumer, anti-discrimination, state
records, critical infrastructure and
cyber security).
Ensure AI system development
complies with the NSW AI ethics
Management
policy, AI assessment framework, Executive level
Level
Product Owners -
organisational values, and related
standards.
Ensure that high-risk AI projects and
Management
solutions are presented to the AI Executive level
Level
Product Owners -
Review Board (AIRC).
Approve AI project and solution risk
Management Executive
tiering, treatment plans and accept Executive level
Level
Product Owners
level
residual risks.
Establish clear data governance
Management
policies for AI systems, including Executive level
Level
Product Owners Everyone
data collection, storage, and usage
Regularly review agency governance,
risk, and compliance frameworks to
Management Executive
ensure alignment with the AIAF and Executive level
Level
Product Owners
level
emerging AI regulatory guidance and
legislation.
Evaluate vendors and third-party AI
solutions for compliance with the
NSW AI Ethics Policy, AIAF, and AI Management Level Product Owners - Everyone
procurement guidance and
framework
Ensure compliance with Digital NSW,
department, and agency policies and
guidelines on using public, non- Management Executive
Everyone Everyone
secure applications, such as Level level
generative AI chatbots (e.g.,
ChatGPT).
Understanding Responsibilities in AI Practices 6
Establish Oversight Mechanisms
ACTION ACCOUNTABLE RESPONSIBLE CONSULTED INFORMED
Create a multidisciplinary AI
advisory board or committee to
Management
monitor and advise AI projects & Executive level
Level
- Everyone
solutions (ethics, legal, technology,
data, privacy etc.)
Include external experts and
stakeholders in Governance,
Management
Assurance, Audit, and advisory Executive level
Level
- -
committees to ensure diverse
perspectives.
Designate a responsible owner for AI Management
Executive level Executive level -
governance in the C-suite. Level
Ensure regular independent reviews
of AI governance and assurance Management
Executive level - Everyone
functions to assess performance Level
and effectiveness.
Ensure that AI systems augment,
rather than replace, human Management
Executive level Users Users
decision-making where its use could Level
create harm.
Ensure AI solutions with medium or
higher risk have incident response
Management Product
plans, with tested, monitored, and Executive level
Level Owners
-
communicated appeal processes
that include human intervention.
Ensure high-risk AI solutions can
provide clear explanations for their
outputs when required and have Management
Product Owners Users Users
established mechanisms to trace AI Level
decisions back to their source data
and logic.
Ensure the use of AI solution Management
Product Owners Users -
benefits outweigh the risks Level
Conduct audits at a frequency
determined by potential risk to
Management Management Product
ensure AI systems meet data quality Level Level Owners
-
standards, desired outcomes, and
NSW ethical policy.
Understanding Responsibilities in AI Practices 7
4 What’s next?
By following this guidance, government agencies can better structure their approach to responsible
AI, ensuring that all levels of the organisation are aligned and actively contributing to ethical AI
practices.
Agencies should start by ensuring a Governance and Assurance function has clear accountability
for overseeing responsible AI use.
Understanding Responsibilities in AI Practices 8
