On light duties

If big tech is known for moving fast and breaking things, government is known for moving slowly and not fixing them. But the third quarter of 2026 has seen a flurry of action in the media regulation space, with the news bargaining incentive, stage 2 privacy reforms, and now the digital duty of care. Still, the question of whether this flurry will fix things will likely take some time to answer. 

The digital duty of care was first announced in November 2024 in response to the statutory review of the Online Safety Act conducted by Delia Rickard, but a full response to the review did not arrive until April of this year. The Rickard report recommended that a duty be imposed on online services ‘to take reasonable steps to address and prevent foreseeable harms on their services’. Importantly, such a duty should be focused on platform systems, such as their algorithms or data-collection practices, rather than on the content they host. This targets regulation at what platforms are clearly responsible for and seeks to prevent harms before they occur, rather than reacting to harms after the fact. It can also help avoid the thorny problems of freedom of expression that accompany obligations to remove or otherwise moderate content.  

A broad, general duty lessens the need to impose specific rules for different kinds of harm and can be more flexible to changing circumstances as a result. This makes a duty potentially suitable as the foundation of a comprehensive digital platform accountability regime and would bring Australia closer to recent global developments in systems-based regulation. But much hangs on how the duty is framed.

The exposure draft frames the duty as a requirement to ‘ensure, so far as is reasonably practicable, a safe online environment.’ What is reasonably practicable is, in part, determined by assessing the risk of harm and the costs and benefits of available ways of eliminating or minimising the risk. This seems broad, but the bill’s specification of ‘safe online environment’ narrows the duty markedly. For children, a safe online environment is one in which they are protected from material and conduct that is harmful to them, and from harms ‘associated with the operation of design features’ of the services. What is meant by harmful material and conduct and by harmful design features is stipulated in the bill, with the latter including recommender systems, endless feeds, feedback mechanisms such as engagement counts, or time-limited content. For adults, a safe online environment is one in which they are protected from ‘seriously harmful material and conduct’ such as that relating to child sexual exploitation, sexual violence, physical violence, harassment or encouraging suicide. There is no duty to protect adults from harms associated with harmful design features. 

The scope of the duty is therefore determined by a legislated list of specific types of harmful material and conduct—wholly so for adults and in part for children. The duty to ‘protect’ users from this material and conduct is likely to encourage a continued focus on content takedown as the most reasonably practicable measure that platforms can take, undermining the new systems focus. Similarly, for system design, there seems little room for nuance, with features such as recommender systems being deemed harmful outright and platforms obliged to remove them for under 16s. The most practicable solution for 16–18-year-olds is likely to be the same, while for adults these features are out of scope. 

Other obligations, however, are broader. To comply with the duty, online services must ‘manage design features appropriately’. This includes through providing ‘user-empowerment tools’, which may be specified by legislative instrument. This provides some room to address system design risks for adults, and the government has already announced plans to specify one such tool—an opt-out mechanism for recommender algorithms. But it seems a cumbersome and relatively inflexible mechanism and is arguably already addressed by the general obligation to effectively manage risks, given suitable guidelines on best practice.  

This obligation could conceivably extend to any type of risk. But given the narrow meaning of ‘safe online environment’, framed in terms of protection from harmful content and conduct, there is little to suggest that it would extend in practice to risks to democracy, electoral integrity, media freedom, or fundamental rights—including freedom of expression—despite the widespread acknowledgement that platform systems contribute to these risks. These risks play a central role in comparable regulation overseas, such as the EU’s Digital Services Act.  

But with conservative parties seemingly having decided to take a freedom-of-speech angle against even the narrow protections provided in this bill, that seems very unlikely. 

References 
Exposure Draft—Online Safety Amendment (Digital Duty of Care) Bill 2026 
https://www.infrastructure.gov.au/department/media/publications/exposure-draft-online-safety-amendment-digital-duty-care-bill-2026 

Rob Nicholls, ‘Australians may soon be able to switch off social media algorithms. Is that censorship?’ The Conversation, September 9, 2026. 
https://theconversation.com/australians-may-soon-be-able-to-switch-off-social-media-algorithms-is-that-censorship-291359