Indonesian government encourages ethical AI to tackle deepfakes  

A deepfake video showing President Joko Widodo giving a speech in Mandarin and Arabic shocked the public last October. In fact, the President cannot speak these foreign languages. This video, amongst others, raised fears about the impact of deepfakes on the general elections held in February.  

Deepfakes refer to AI-created videos that are very convincing and have the potential to mislead viewers, create hoaxes, and perpetrate frauds.  

"Based on the patterns [we observed] in Indonesia, the deepfakes mostly use faces of real people. The voice is similar, but what is talked about does not match the character of the person," Usman Kansong, Director General of Public Information and Communication at the Ministry of Communication and Information Technology (Kominfo), tells GovInsider.  

He reveals that ahead of the February general elections in Indonesia, several deepfakes targeting political figures were being circulated. Apart from President Jokowi, visual materials of polical party leader Surya Paloh and Anies Baswedan, a presidential candidate, were also misused.  

Furthermore, a deepfake of the ex-Minister of Health went viral, in which he seemed to be promoting a drug that was not registered with the Food and Drug Administration (BPOM). 

Minister issues regulatory letter to anticipate deepfakes 

To tackle deepfake crimes, Minister of Kominfo Budi Arie Setiadi has issued a regulatory letter on the ethical use of AI, addressed to business actors in the public and private spheres. 

First, AI should be used to support human activities, especially to increase user creativity in solving problems and work. Second, the use of AI must maintain privacy and data, so that no individual is harmed. Third, there should be supervision of AI use to prevent misuse by the government, tech providers, and users.  

"Kominfo encourages companies and organizations that develop and use AI to pay attention to the principles in the ethical guidelines," Usman said. For example, Google has implemented watermark technology, along with other players such as Adobe and Intel, to highlight when media has been digitally altered. 

In addition, Kominfo continues to encourage the Press Council to prioritize the ethical use of AI in journalism.

The Indonesian Public Relations Association (Perhumas) is currently creating ethical guidelines, which are planned to be launched in November. Until now, apart from Kominfo, only the Financial Services Authority (OJK) and FinTech associations have released ethical guidelines for using AI.

OJK released its code of conduct for the responsible use of AI in FinTech in early 2024. For example, fintech providers must have a liability framework for the output generated from AI-based applications, to provide accountability to consumers.  

AI-powered machine to identify harmful content

Kominfo shares that there is no specific technology to identify deepfakes. "We are encouraging technology manufacturers to prevent deepfakes," Usman said. He said that Kominfo had also met all social media platform organizers ahead of the last election, to increase vigilance for such content. 

The Ministry said there were 3,235 election-related hoaxes circulating on social media platforms and search engines since July 17, 2023. Of that number, 1,971 of them have been taken down. 

Usman explained that since 2018 Kominfo has utilised “AIS”, an AI-based crawling machine, to identify harmful content on the Internet, including hate speech, pornography, fake news, and threats.   

The technology classifies up to millions of links detected with harmful content. Kominfo then orders the content provider to remove the harmful content within 2x24 hours. 

Kominfo also cooperates with the Directorate of Cyber Crime of the National Police to eradicate harmful content.  

"The National Police can also request information from Kominfo if there are findings," Usman added.  

In addition, Kominfo actively processes reports from the public to help officers identify harmful content more quickly, considering that there is no single technology that is 100% capable of detecting hoaxes. 

Currently, Indonesia does not have specific government regulations regarding deepfakes and any form of image or video manipulation by AI.  

The creation of new content edited from the original content with the intention of spreading false news is categorised as a hoax under current laws. Hoaxes are considered complaints, so the perpetrators cannot be prosecuted if there is no report of a criminal offense or loss.  

This article was originally published in Bahasa Indonesia.