MANILA — The release of a video purportedly showing President Ferdinand Marcos Jr. doing drugs that the Philippine National Police and National Bureau of Investigation have said is fake could lead to an escalation of the use of deepfakes, an AI expert said. 

The Department of Information and Communications and Technology has also dismissed the video, which was shown in gatherings of the Duterte-affiliated Maisug, as a "a poorly designed" fake video. In April, a manipulated audio clip of Marcos Jr allegedly authorizing the use of force in the West Philippine Sea also spread online.

READ: Officials dismiss 'fake, malicious' video showing Marcos allegedly doing drugs

"I think gantihan na (I think there will be a tit for tat). First, to debunk the 'polvoron' video, then later to put other people in harm's way," Dominic Ligot of Data Ethics PH told ABS-CBN News in an online exchange.

"It seems [Duterte supporters] fired the first shot. Arms escalation na 'yan (This will be an arms escalation)."

He added that the deepfake video and audio are "easier to do now" because of developments in technology and that these "will get better and better."

A Taiwan FactCheck Center report in February noted that the AI tools to make deepfakes are readily available and are easy to use.

Government agencies were quick to dismiss the video, which went viral on the day of President Ferdinand Marcos Jr.'s State of the Nation Address, as fake and have warned of potential legal cases against the people behind them.

Former President Rodrigo Duterte, who previously accused Marcos Jr. of drug use, said he had nothing to do with the release of the video but also said the President must take a drug test to dispel doubts. 

Duterte had refused a similar test amid rumors of fentanyl use when he was president.

RELATED: Cybersecurity experts warn public vs 'deepfake' video boom

'NOT NECESSARILY TO CONVINCE'

Experts said disinformation does not necessarily need to convince the public but can be used to cast doubt on and discredit their targets.

"Parang De Lima lang (Like what they did with De Lima) and that was even worse," he said, referring to a supposed sex video that House leaders claimed in 2016 to have seen and planned to show in legislative hearings into her since disproven involvement in the illegal drug trade.

Disinformation researcher Jonathan Corpuz Ong, in a separate online exchange, said that disinformation and manipulated images "can be used to revive old narratives and rumors, which this deepfake appears to do."

Ong, who is director of the Global Technology Lab for Social Justice and is a professor at the University of Massachusetts Amherst, added the video is "giving new life to old rumors in light of Marcos positive momentum." 

Creators of deepfakes may have another motive, he said: "In some other cases the audience of the video isn't the general public at all but politicians and potential clients.

In those cases, videos could be created as "mock ups and proofs of concept for future campaigns."

DEEPFAKES IN ELECTIONS IN ASIA

Deepfakes, or 'cheapfakes' for the badly-done ones, have figured in politics elsewhere in Asia.

Voters in India, which held its general elections from April to June, had to contend with "fake audio messages to funny memes and videos of politicians mostly on the lines of trolling their rivals," according to Ruby Dhingra, managing editor of fact-check organization Newschecker.

"More concerning were the AI-manipulated videos of Bollywood superstars endorsing or criticizing specific political parties," she told ABS-CBN News in an online exchange.

She said that criticism or endorsement that is supposedly from a celebrity "can significantly impact voting behavior" especially since "most consume social media passively and do not critically evaluate whether something is fake."

Sometimes, she said, deepfakes are released "to create mischief, sow the seeds of doubt, and make people question what they see and hear, reducing overall trust." 

Eroded trust, she said, could make the public easier to manipulate.

RELATED: YouTube tightens rules on AI, deepfakes, political ads amid global elections

Fact checkers like Dhingra use reverse image search, check for red flags like lips not moving in sync with words, and look for the original source video to check videos that are suspected to be deepfakes. 

Taiwan, which had its presidential elections in January, also had to deal with deepfakes, including a supposed voice recording of Democratic Progressive Party Lai Ching-te — since elected president — calling himself immoral and calling the party "fraught with scandals," according to the Taiwan FactCheck Center.

"The voice in the video indeed sounded like Lai's voice, but it was absurd that Lai would make such a comment… There was no other media coverage of Lai's outlandish remark in the video," TFCC research fellow Dr. Wei-Ping Li said in a February report.

The proliferation of deepfakes "highlight a disturbing trend in how AI will exacerbate the disinformation problem," Li said, especially with tools that "can create an AI film of nice quality in less than thirty minutes."

Claire Chen, deputy editor-in-chief of the Taiwan Fact Check Center, said they use AI detection tools and interviews with AI experts in their fact checks of deepfakes.

"[We also] use traditional fact-checking tools like reverse image search to find the original video," she said in an online exchange. 

"You can use Hive Moderation to detect AI in video," she also said, referring to a paid service that offers a demo on its website.

She said there is not enough data yet to say how deepfakes and other forms of disinformation affected Taiwan's elections, "but we need to let the public know of the danger."

WHAT CAN BE DONE?

Li said that media literacy and fact-checking can help counter disinformation and the Taiwanese government has used laws against spreading false information to prosecute people engaged in disinformation.

"However, in practice, it would be hard to locate the creators and spreaders of the false information, not to mention produce evidence of the actors' intentions," she also said.

She said the speed and proliferation of these deepfakes highlights the need "to develop more tools" to address the problem of disinformation.

Ong has long called for a multisectoral and  multipronged  approach to disinformation, including recognizing that manipulated information may be resonating with valid frustrations or concerns.

In a separate but related commentary on a US disinformation operation against Chinese vaccine Sinovac in countries, including the Philippines, he called on independent researchers to help look into digital operators in the Philippines "who actually executed the campaign and translated them to local audiences in the precise online communities that allowed these stories to take root."

RELATED: DICT says no to TikTok ban, but yes to regulating apps

He added a civilian-led inquiry could also call attention to "how social media platforms should be more transparent and accountable especially to victims of disinformation campaigns in the Global South."

Ligot said that media and those working to address misinformation and disinformation should increase awareness of deepfakes "what kinds there are and how they work."

He said this would help make the public more suspicious of the content that they consume and share.

In the meantime, the Department of the Interior and Local Government has ordered the Philippine National Police to conduct an investigation into who created the video, with the Department of Defense and the Armed Forces of the Philippines saying they are willing to help in any inquiry.