EXPLAINER: What are deepfakes?

The Presidential Communications Office (PCO) recently flagged a video featuring a deepfake audio of President Ferdinand “Bongbong” Marcos Jr. supposedly instructing the military to act against a foreign country.

Marcos never gave such an order, the PCO said, and that his voice was mimicked in the video through the use of generative artificial intelligence (AI).

“Ginamit lang 'yung boses ng Pangulo at ginawan siya ng script para ilapat doon sa video na mukha siyang mismo ang nagsasalita. Nililinaw lang po ng PCO na hindi ito totoo,” PCO Assistant Secretary Dale de Vera told a public briefing on Wednesday.

(They spliced the voice of the President and used it for a script of their video. We would like to clarify that it is not true.)

The video circulating online has since been deleted, and the channel that uploaded it was banned by YouTube, but not without raising concerns.

National Security Council (NSC) Assistant Director General Jonathan Malaya warned that the video could be used to muddle the issue on the West Philippine Sea, such as the purported "gentleman's agreement" regarding the Ayungin Shoal.

"Pinagsabong yung mga Pilipino, pinagsabong tayong lahat eh, ngayon nagtuturuan na. Sino 'yung ganito sino 'yung ganyan, then we saw there was a foreign hand that was manipulating, some form of information manipulation," said Malaya.

(They are pitting Filipinos against each other. We saw there was a foreign hand that was manipulating, some form of information manipulation.)

Deepfakes, according to the Department of Information and Communications Technology (DICT), are the “mimicry of the voice or image of the person purporting to be someone else.''

It can be generated by splicing photos, videos, audio, or a combination of the formats to create deceiving and false content.

“Deepfake technology is a tool being used either on its own or with the use of AI…to mimic the imagery and the voice, or sometimes both, of a particular individual by other persons,” DICT spokesperson Assistant Secretary Renato Paraiso told GMA News Online.

“This can be done in real-time or as recorded,” he added.

Now, how does AI play a role in its proliferation? Generative AI is making it easier and faster to create deepfakes, said Paraiso.

“You just program it tapos AI na ang kumukuha through all available sources. Whether it's on the internet or social media pages, there are other sources, sila ang kumukuha ng mga sample ng mga voices. 'Yung intonation, especially 'pag may mga words kayo in-utter pinagtatahi-tahi niya po para magaya ang boses,” he added.

(You just program it and then AI will gather all the available sources. Whether it's on the internet, social media pages, or other sources. It stitches the spliced materials in order to imitate the voice.)

“With the imagery, it's the same thing. It gathers imagery, the pixelated imagery…Even in real-time puwede siya mag-deepfake. Ganun ka-advance ang technology ng deepfake para magaya 'yung image at mannerisms at galaw ng isang tao,” he added.

(It’s the same thing with images. It can even do it in real time. That’s how advanced the technology is.)

Paraiso said people use deepfakes in several ways. Some users utilize them to create parody content, while others were reported to have been crafting deepfakes for “nefarious” intent.

“Yung more nakakabahalang use ng deepfake is to sow disinformation and distrust sa ating kababayan and ‘yung recently na in-attempt na magkaroon ng civil unrest, and it delves into the national security of our country,” he said.

(The more nefarious use of it is if it is being used to sow disinformation and distrust. Recently, it was also attempted to create civil unrest, and it delves into the national security of our country.)

“Wala akong magandang nakikitang idinudulot at kapakinabangan sa deepfake. 'Yun nga yung example natin kanina na pang-katatawanan lang, still, it causes confusion,” the DICT official added.

(I don’t see any good use for it. Even parody content causes confusion.)

In 2019, a video was edited to make it appear that former senator Leila de Lima had acknowledged that she was a protector of the Bilibid drug lords and that she would step down from office.

Several global personalities have also fallen victim to deepfakes, including Pope Francis, US President Joe Biden, and Taylor Swift.

"The technology of simulation behind these programs can be useful in certain specific fields, but it becomes perverse when it distorts our relationships with others and with reality," the pope wrote in January.

Deepfake regulation

With the advent of deepfake technology, how is the Philippines addressing it? Paraiso said the country still lacked laws for its regulation, but there are measures penalizing erroneous use, including spreading fake news.

The DICT cited Article 154 of the Revised Penal Code, which states that “any person who, by means of printing, lithography, or any other means of publication, shall publish or cause to be published as news any false news which may endanger the public order or cause damage to the interest or credit of the State.''

It is punishable by arresto mayor or jail time of one month and one day to six months and a fine.

“Deepfake is a crime of content. Hindi naman ibig sabihin 'yung mismong technology is bad or the use of AI is bad pero that particular law, ang kelangan natin ma-address yung content creators nito kung sino ang gumagawa nito,” said Paraiso.

(Deepfake is a crime of content. The use of deepfake or AI is not bad, but what we need to address is the content creators involved in this.)

In February, the DICT appealed to legislators to craft laws for AI regulation as the 2025 midterm elections draw near.

Currently, the DICT is coordinating with social media platforms such as Meta, X, and TikTok to curb malicious deepfakes in the country. It also signed a memorandum of understanding (MOU) with the sites last year to fight fake news.

“We signed an MOU last year wherein talagang sinasaad dito na makikipagtulungan sila with the government's effort to prevent this particular fake news and to tailor fit their community guidelines and policies para ma-address 'yung challenges na kinakaharap natin, specifically dun sa social circle natin sa Pilipinas,” he said.

(We signed an MOU last year and it states that they will coordinate with the government's effort to prevent this particular fake news and tailor their community guidelines and policies to address our current challenges.)

“’Yung kasi bawat bansa iba-iba 'yung tradition, customs. Kailangan 'yung mga kinakaharap natin na challenges, kailangan ma-tailor fit at ma-address 'yung challenges na 'to ng mga social media platforms,” he added.

(Every country has its traditions and customs. Social media platforms need to tailor their community guidelines to address challenges.)

Meanwhile, the DICT issued reminders to the public to prevent falling victim to deepfakes.

“To further prevent 'yung pagkalat o paniniwala sa deepfakes, only go to trusted sites, news forums. Kasi ito yung mga nakikita natin sa social media, hindi natin ma-verify. Hindi natin makita kung saan talaga 'yung sources po nito,” said Paraiso.

(To further prevent falling victim to deep fakes, only go to trusted sites and news forums. We cannot verify what we’re seeing online since we don’t know its sources.)

“Maging mapagmatyag tayo. Maging maalam sa mga emerging new technology,” he added.

(Let us be vigilant, especially with emerging technologies.)  — VBL, GMA Integrated News