ACS calls on government to embed people into AI safety plan

The Australian Computer Society (ACS) has warned the federal government that legislation and safety frameworks alone cannot protect Australians from AI-related risks.

Responding to the government’s six proposed AI safety protections and the 18 recommendations stemming from the Rapid Review into the OpenAI Medicare debacle, ACS CEO Dr Pins Ralston said capable technology professionals will be needed to effectively implement AI guardrails.

“Legislation and frameworks only work when capable, trusted technology professionals put them into practice,” he said. “Where are those people in the plan? How will we know they have the capability to do the job?

“The review calls for stronger AI literacy, which can help staff recognise and report a problem. When an AI agent gains access to a government system it shouldn’t, someone has to establish what it accessed, whether other public services are at risk and how quickly to act. That requires more than mere literacy: it takes specialist resident AI and cybersecurity capability.”

Ralston said the same is true of independent testing, with experts needed to put a powerful AI model through its paces, uncover risks and challenge the model developer’s safety claims.

The ACS is calling on the government to set clear capability requirements for the people tasked with independently testing high-risk AI models and investigating serious AI incidents. Capabilities should be assessed against recognised skills frameworks such as the global Skills Framework for the Information Age, and for high-impact state-funded uses of AI, a capable resident technology professional should be appointed as a senior owner and given authorisation to pause or withdraw use of the model, the ACS recommended.

For large-scale AI training, a qualified compliance officer based in Australia should be required, alongside notification and defined capability evaluations, the society added. Meanwhile, the APS workforce should be equipped with professional AI capability that goes beyond literacy, the ACS said.

Research from MIT and the University of Queensland indicates that only 23% of Australians trust AI developers to ensure their software is safe.

The six proposed AI safety protections would require frontier AI developers setting up shop in Australia to maintain security controls across all stages of development and deployment, identify and manage serious risks associated with their models, and allow independent evaluation of their models. The developers would also be required to provide mandatory notification of incidents to the government rapidly, maintain and publish safety and security frameworks, and respond to updated standards and rules as technologies change.

Under the proposal, a targeted regulator for AI safety would be established, which would have the ability to impose penalties on companies that fail to comply with the mandatory obligations.

Announcing the proposed obligations, Prime Minister Anthony Albanese said AI presents enormous opportunities for Australia, but also new risks.

“These protections will ensure we keep Australians safe and secure our national interests, while seizing the generational opportunities that AI offers,” he said. “Our world-leading AI in Australia’s Interests agenda aims to provide Australians confidence that frontier AI companies are taking steps to improve safety and security, and provide certainty for investors — boosting productivity, building economic resilience, while bolstering sovereignty.”

Image credit: iStock.com/cherdchai chawienghong