Why Australia’s fight with X shows platform neutrality is a myth

2026-04-22 Tianming Fang Uncategorized 0

Social media companies can easily describe themselves as neutral. They like to describe themselves as a neutral space: people publish content, debate, joke and debate here, while the platform itself just stays out of the way and allows freedom of speech to develop. However, when hate content is openly spread, security teams are cut down, and those who become targets can only report, record and bear the harm themselves, this statement is difficult to be convincing.

Network hate speech is not only a problem of “bad users’ bad speech”, but also a problem of platform governance. The platform determines which content will be seen, which content will be streamed, and which content will be deleted. They also decided to give users the right to report, but it is also the only right for users to rely on the system when they are injured. Therefore, the problem lies not only in hate speech itself, but also in the systems that help it spread and keep it visible.

What happened between Australia and X?

The dispute between Australia and X did not arise overnight. In June 2023, the Australian cybersecurity Commissioner issued a legal notice to X in accordance with the Online Safety Act, requiring the company to explain what measures it has taken to deal with cyber hate and implement its own hate behavior policy. In January, 2024, the cybersecurity officer released the summary of X’s response, and said the platform reduced trust and security personnel, slowed down the response speed to network hate reports, and re-enabled thousands of previously banned accounts, but did not further review them.

Video 1. ABC News report on Australia’s eSafety regulator criticising X over its handling of online hate (10 January 2024).

These details show that network security does not just about whether a bad post is eventually deleted. It was influenced by staffing decisions, resource allocation, internal processes and platform priorities at an earlier stage. Once the platform weakens its security system and reduces the number of people responsible for managing risks, harmful content is more likely to spread before any measures are taken.

This is also where Sarah T. Roberts’ research on content auditing has become useful. Roberts believes that most major platforms follow the logic of “publishing content first and then auditing”: content is published immediately, while auditing usually takes place later. From the perspective of growth and participation, this may be reasonable, but it also means that the harm often spreads before someone intervenes. Therefore, the problem is not just audit failure. The problem is that the platform model itself does not give priority to these factors in terms of security.

Full-Report-Basic-Online-Safety-Expectations-Summary-of-response-from-X-CorpTwitter-to-eSafetys-transparency-notice-on-online hateDownload

File 1. eSafety’s summary of changes to X’s trust and safety resources, January 2024. Source: eSafety Commissioner (2024).

The platform is not just “carrying” speech

The concept of platform neutrality can only be established on the premise that the platform is only responsible for providing the stage. But social media doesn’t work that way. The platform uses recommendation system, popular list, forwarding function and ranking tools to organize attention. This means that they not only carry words, but also affect their visibility.

This is why some scholars believe that the role of the platform is far more than providing people with space to speak. For example, Matamoros fern á ndez (2017) pointed out that the platform can amplify and repeat racist speech through design, business model and governance options. A hate speech spreads not just because someone wrote it. The reason for its spread is that the platform makes the spread fast, smooth and profitable.

If platforms actively can decide what things we can see and how content circulates, Then the neutral argument is untenable. We can find that platforms already intervene all the time. If the platform can proactively decide what content we can see and how it will be disseminated, then this neutrality theory is untenable. We can see that the platform has been intervening. When the platform begins to decide which content will be recommended; When they define what speech constitutes hate speech; When they choose to delete certain posts or choose to ignore other posts; When users often feel that the reporting system is slow and ineffective. They are always intervening. In this sense, platforms are not just “carrying” harmful speech, they can be called managers.

Why Platform often fails to detect hazards

One reason for the repeated failures of platform governance is that not all network hate speech is obvious to the platform. It does not always take the form of direct insults or threats that are clearly illegal. Sometimes it is presented in the form of humor or meaningful only in a specific social context. This is why standardized moderate management often fails to detect the harm.

Platforms are heavily dependent on non-human systems for large-scale management. But hate intention is often related to context, indirectness and culture. As Roberts (2019) has shown, deciding whether content is acceptable often involves more than detecting keywords or obvious images. Intention, context and cultural meaning can all work at the same time, which is why software or algorithms themselves are often not enough. When the construction of Management Systems focuses on scale rather than context, they are more likely to miss subtle, obscure or local harmful remarks.

Sinpeng et al. (2021) put forward a similar view in their research on the regulation of hate speech in the Asia Pacific region. They believe that hate speech is deeply influenced by language and context, which means that the norms and standards of the global society and automatic classifiers are often unable to accurately capture the point where the target group is attacked by hate speech.  A post might seem vague or harmless to a moderation system, but the people being targeted often know exactly what it means.  This is because online hate speech is often expressed through language, jokes or local culture, and outsiders including the platform may not fully understand it. In other words, systems built for large-scale applications are often difficult to identify those subtle, obscure and contextual injuries.

When the platform fails to identify and respond to hazards, the burden of processing falls on users.

This is well reflected by the conflict with the X platform. Those who become the targets of attacks often need to pay attention to hate speech and record it, report it and explain its harmfulness, and then wait for the platform to respond. In this sense, the reporting system of X platform not only provides support, but also may transfer the responsibility from the platform to those who suffer malicious attacks.

In the summary report in January 2024, e-safety said that the X platform has cut trust and safety staff, responded more slowly to reports of online hate speech, and has reactivated thousands of previously banned accounts without additional review. They show that while users are required to rely on the reporting system for protection, the X platform weakens its security capabilities.

From this perspective, the problem is not just about hate speech still exists online. It‘s about the X platform that makes users undertake too much actual and emotional audit work. People need to identify hazards, propose hazards to X and explain hazards, then wait for a long time, while the platform continues to label itself as a neutral intermediary. The dispute between Australia and X reveals that when the security capability of the platform declines, protection will become slow and weak, and more dependent on those who have been attacked to report themselves.

Australia has had such problems before. For a long time, the Australian network platform has been difficult to deal with racist harm on the network fairly. As Matamoros fern á ndez (2017) and Carlson and Frazer (2018) have expressed, the problem is not only that harmful content appears on the Internet, but also that the platform system will magnify the harm and put too much responsibility on the targeted victims. In this sense, the problem of the X platform is not a new one, but rather an old one that reappears in a slightly different form.

Why is the responsibility to build a safer system more meaningful

If the problem of the X platform is that hate speech stays online for too long, the solution seems to be very simple: hire more auditors and speed up the deletion of content. But the problem goes far beyond that. The formation of online hate speech is affected by platform design, visibility system, reporting process and audit inequality, which already exist before any single post is audited. That’s why it’s always hard to keep up with the situation if you focus on a single content model. As Woods and Perrin (2022) pointed out, cyber harm should be regarded as a systemic problem, not just a content problem.

Therefore, the responsibility to build a more secure system becomes crucial. This means that the platform should take reasonable measures to reduce the foreseeable harm, rather than waiting for users to report after being attacked. This does not mean that platforms should be responsible for every piece of harmful content, but they should be responsible for the way their systems are built and the increased risks of these systems.Australia is already moving in this direction. In November 2024, the Australian government said that it would enact the  Digital duty of care to make the platform bear more responsibilities to prevent network injury and reduce foreseeable risks. Subsequently, the government reiterated this practice in its report on Online safety act 2021.

The key here is the change of responsibility. It is not only about whether harmful content has been deleted after the platform, but also whether the platform has taken measures in terms of system design, security personnel allocation and complaint process organization to reduce the possibility of harm from the beginning.

Conclusion

The dispute between Australia and X makes people have to admit the fact that the platform is not just a neutral space that simply carries what users say. They will guide the direction of information dissemination, determine what content can be presented continuously, and who will be responsible for handling the resulting injury in case of safety system failure. This is why cyber hatred cannot be regarded as a single problem of bad speech or bad users. It is also a problem of platform governance.

Reference list

Be the first to comment

Leave a Reply

Your email address will not be published.


*