Academic highlights need for PSE accountability over AI-generated information

Jakarta (ANTARA) – Kunto Adi Wibowo, an academic at the Faculty of Communication Sciences at Padjadjaran University (Unpad), has highlighted the urgent need for electronic system operators (PSEs) to provide accurate information to people using artificial intelligence (AI).

The head of Unpad’s Doctoral Programme in Communication Sciences made the remarks amid widespread online speculation that the government was restricting information about forest and land fires, following information provided by Meta AI on Threads.

“Meta AI is not an absolute source of truth, and its answers are not official statements from Meta. Generative AI technology still has limitations and can produce inaccurate information. If a machine says the government requested content restrictions, the question is simple: what is the evidence? Where are the documents or the basis for the decision that can be verified?” Kunto said in a statement received in Jakarta on Friday.

Kunto said Meta, as a PSE, could not avoid responsibility when its AI product generated unverified information, particularly when it concerned government policy and people’s freedom of expression.

He said statements by the government and information from technology companies presented by AI should be testable against evidence in any scenario.

The public should also remember that answers generated by AI services such as Meta AI cannot be treated as official confirmation from the company, let alone as the sole basis for accusing an institution of restricting information.

This is because AI can suffer from what is commonly known as an “AI hallucination”: it may provide users with answers that sound convincing despite lacking factual support.

The risk is greater when users assume AI-generated answers are official information from the company that developed the system.

Kunto said the issue had emerged in connection with Meta AI on Threads and allegations of restrictions on information about forest and land fires.

In these circumstances, he said, Meta needed to act responsibly and demonstrate transparency about its algorithms. It was not enough for the company simply to provide an AI product; it also had to ensure users understood its limitations.

Similar AI hallucinations on Meta’s services have previously occurred in the company’s home country, the United States. One example involved the mistaken suspension of the Instagram account of No Kings Coalition, a group associated with demonstrations against the government of President Donald Trump.

Meta Vice President Andy Stone quickly apologised for the error. Meta acknowledged that the account should not have been suspended and restored it, although restoring access took a long time.

For Kunto, the incident showed that decisions made by Meta’s automated systems were not always accurate and could directly affect public perceptions and freedom of expression.

Although the cause may not be the same, he said, the US experience was grounds for demanding that Meta be open about the alleged content restrictions in Indonesia.

“In the United States, Meta itself has acknowledged that its systems can make mistakes, causing accounts that should not have been blocked to lose access. This means we cannot simply assume that all Meta moderation decisions are correct. The company must be able to explain and take responsibility for its system’s decisions,” Kunto said.

Regarding allegations that forest and land fire content had been removed in Indonesia, Kunto questioned whether Meta’s automated moderation system was sufficiently capable of understanding the context of images and videos posted by the public.

He cited footage of forest fires showing flames, smoke, crowds of residents or emergency responders at the scene. He said it should be examined whether Meta’s moderation system might mistakenly classify such images as violent content, riots or violations of community guidelines.

“Could images of flames and smoke during forest and land fires be misread by a machine as riot content? Could there be a misclassification like the one that occurred in account moderation in the United States? We do not know. Meta is the one that must explain how its system works and why the content was restricted,” he said.

Kunto said documentation of forest and land fires was in the public interest. Information about fire locations, smoke spread and their effects on communities was needed to raise awareness and help manage disasters.

He stressed that Meta, as the platform operator, should provide the explanation. Without transparency, the public would continue to speculate about why content had disappeared, while the technology platform making the decisions would feel no need to take responsibility.

“If there was a government request, Meta must be able to show its legal basis. If content was removed because of an algorithmic error, admit it and fix it. The public should not end up accusing the government while the company controlling the platform provides no explanation,” he said.

Kunto also recommended that Meta improve transparency by adding labels or further explanations about the accuracy of its AI.

He noted that Meta’s competitors, OpenAI’s ChatGPT and Google’s Gemini, both tell users upfront that their AI may not be entirely accurate and recommend that they verify the information provided.

Kunto said an upfront explanation on Meta AI would be crucial to prevent system errors from creating further misinformation and limiting people’s ability to make informed decisions.

He concluded that the considerable influence of digital platforms such as Meta must be matched by a corresponding level of responsibility in the countries where they operate.

Meta not only controls the distribution of information through its algorithms and moderation policies, but also provides AI technology whose answers can influence public understanding and opinion about events.

“Meta has considerable power to determine what information the public can see. Such power must come with transparency and accountability. Technology companies must not decide what the public is allowed to see and then refuse to take responsibility when problems arise over information the public needs,” Kunto said.

Meta has so far said it plans to improve transparency about information accuracy by adding a disclaimer to every piece of content shared by Meta AI. The requirement may take effect from 30 October 2026, with the explanation: “Meta AI can make mistakes and customers are responsible for the consequences of using Meta AI.”