Sibenco Legal & Advisory

AI in Australia’s Interests Announcement

gvnet

by

Australia’s national-interest approach is emerging as other major jurisdictions recalibrate AI security, innovation and regulation.

On 15 July 2026, Prime Minister Anthony Albanese announced that the Australian Government would establish an Office of AI within the Department of the Prime Minister and Cabinet and develop a set of Australian Standards for AI. The announcement formed part of the Prime Minister’s speech, AI in Australia’s interests, which emphasised sovereign capability, national security, economic resilience, infrastructure, copyright and ensuring that AI investment occurs on Australian terms.

Although described broadly as Australian Standards for AI, the announced requirements are directed principally to large AI data centres and the infrastructure supporting large-scale AI training. They are intended to build on the Government’s existing Data Centre Expectations and address power generation, grid connection costs, energy and water efficiency, location, approvals and the conditions under which major AI investment occurs in Australia.

The Government describes the standards as a nationally consistent regulatory framework for large data centres and AI training. The new Office of AI will coordinate implementation across government, while the proposed approach is to be considered by National Cabinet before legislation is introduced. The announcement therefore signals a more interventionist and centrally coordinated approach, but it does not yet establish a comprehensive regulatory framework governing all AI developers, deployers or high-risk uses.

Australia’s existing AI governance arrangements remain important. The Essential AI practices(opens in new tab) contained in the Australian Government’s Guidance for AI adoption, continue to provide the principal responsible-governance benchmark for organisations developing and using AI. They address accountability, impacts, AI-specific risk management, transparency, testing and monitoring, and meaningful human oversight. Commonwealth agencies are also subject to mandatory requirements under the Policy for the responsible use of AI in government.

The proposed Australian Standards appear likely to build on this existing architecture and may be accompanied by further consumer-safety or other regulatory measures. However, their broader scope and relationship with existing guidance, laws and sector regulators remain to be clarified.

Read more: Australian Government: AI in Australia’s interests(opens in new tab)  |  AI in Australia’s interests | Prime Minister of Australia(opens in new tab)

by

What organisations can learn from the Privacy Commissioner’s decision to conclude preliminary inquiries without further action.

The Privacy Commissioner has completed preliminary inquiries into the 2025 Qantas data incident, which affected approximately 5.12 million Australians following a social-engineering attack on an overseas third-party contact centre provider. After almost a year of inquiries, the Commissioner concluded that the evidence did not reveal omissions or failings that made a contravention of the Privacy Act 1988 (Cth) sufficiently likely to justify a full investigation or further action at this stage.

The finding does not suggest that the scale or impact of the breach was unimportant. It demonstrates that regulatory defensibility depends on evidence of the reasonable steps taken before an incident and the quality of the response when it occurs.

What supported Qantas’s position?

The OAIC examined Qantas’s compliance with APPs 1, 8 and 11, including its governance arrangements, security controls and oversight of the overseas service provider. Relevant factors included:

  • privacy and cyber security due diligence before appointing the provider;
  • contractual privacy, confidentiality and security obligations;
  • role-based access controls, workforce screening and training;
  • monitoring, alerting and audit arrangements;
  • incident-response processes and escalation pathways;
  • prompt investigation of unusual login activity, revocation of affected access and containment of the platform;
  • engagement of specialist legal and forensic advisers;
  • public disclosure and subsequent communication with affected customers about the categories of information involved; and
  • extensive cooperation with the OAIC, supported by information and documents.

Practical lessons for organisations

The report reinforces several practical steps organisations should take now:

  • Document why security and privacy controls are reasonable and proportionate to the information, systems and threat environment.
  • Treat third-party contact centres, cloud platforms and service providers as part of the organisation’s control environment, with due diligence, enforceable obligations, monitoring and assurance.
  • Ensure alerts are escalated to people with authority to investigate, contain access and activate incident-response arrangements.
  • Retain reliable logs and decision records so the organisation can reconstruct what occurred and demonstrate its response.
  • Prepare communications that are accurate, staged where necessary and responsive to what affected individuals need to know.
  • Review controls after the incident and preserve evidence of remediation, testing and continuing monitoring.

The key takeaway is that regulatory defensibility is built before an incident. Policies alone are insufficient; organisations need evidence that controls operated, risks were overseen and the response was prompt, informed and proportionate.

Read more: Privacy Commissioner’s media statement  |  Report into preliminary inquiries of Qantas

by

Recent incidents show why advanced AI agents require technical containment, least privilege, monitoring and effective human control.

Four disclosures between 24 July and 6 August 2026 have brought the cyber risks of increasingly capable AI agents into sharper focus. On 21 July, OpenAI disclosed that a combination of its models had exploited a previously unknown vulnerability to establish internet connectivity and accessed Hugging Face’s systems while pursuing a cyber benchmark objective, an incident subsequently highlighted by the Australian Signals Directorate.  On 30 July, Anthropic disclosed that models had accessed the systems of three external organisations during testing after being given unintended internet access. On 4 August, the UK AI Security Institute reported that Anthropic and OpenAI agents had taken 19 unsanctioned actions directed at real people and organisations during evaluations conducted between 25 and 28 July. On 5 August, Meta confirmed that one of its models had exploited a vulnerability in an external service after a third-party evaluator inadvertently provided internet access.

The incidents differed in their technical circumstances and occurred in deliberately permissive or misconfigured testing environments. Notably, the Anthropic, Meta and second OpenAI incidents each arose in the evaluation environment of the same third-party testing firm, Irregular, a reminder that outsourcing a function does not outsource responsibility for governing it. The important lesson is not that the systems were ‘rogue’ in a human sense. It is that agents with objectives, tools, network access and operational privileges can identify alternative pathways and act beyond intended boundaries when containment, permissions, monitoring and human control are inadequate. Instructions and model-level safeguards are not sufficient on their own.

ASD guidance for boards

On 5 August 2026, the Australian Signals Directorate and the Australian Institute of Company Directors released Frontier AI cyber threat considerations for boards of directors. The guidance explains that frontier models can identify and weaponise vulnerabilities, combine lower-severity weaknesses into high-impact compromises and perform malicious cyber activity with little or no human oversight. They can also lower the expertise needed to conduct sophisticated attacks and compress vulnerability discovery and exploitation timelines from days to hours.

Boards are urged to reconsider whether existing cyber risk assumptions and tolerances remain valid. The guidance asks boards to challenge management on attack surfaces, vulnerability remediation, legacy systems, identity and access management, third- and fourth-party exposure, incident response, continuity and recovery. It also identifies dependency on AI providers and foreign ownership, control or influence as cyber supply-chain issues.

The continuing relevance of ASD’s agentic AI guidance

ASD’s July guidance on the careful adoption of agentic AI in cyber defence remains directly relevant. It recommends beginning with clearly defined, lower-risk use cases and increasing autonomy only as confidence and assurance mature. Core controls include:

  • minimum necessary permissions and strict isolation between systems and environments;
  • human approval for sensitive, high-impact or irreversible actions;
  • continuous monitoring of agent behaviour, decisions and tool use;
  • comprehensive logging, auditing and accountability mechanisms;
  • red teaming, adversarial testing and security assessment;
  • validation of third-party tools, integrations and dependencies;
  • clear mechanisms to interrupt, halt and, where possible, reverse agent actions; and
  • defence in depth across inputs, data sources, tool integrations, outputs and agent-to-agent communications.

The new frontier-model guidance broadens the issue beyond an organisation’s own use of AI. Even organisations that do not deploy advanced agents may face attackers using them. Boards therefore need visibility of both sides of the risk: the security and governance of internal AI use, and the adequacy of cyber resilience in an environment where adversaries can operate at machine speed.

Read more: Frontier AI cyber threat considerations for boards(opens in new tab)  |  Careful adoption of agentic AI in cyber defence  |  Five Eyes guidance: Careful adoption of agentic AI services(opens in new tab)