Despite the rise of deepfakes, cheap fakes continue to remain a problem, a foreign policy and national security expert of the United States says.
Dr Heather Ashby was sharing how technology impacts elections at an event on Tuesday jointly organized by the US embassy in Dhaka and the Diplomatic Correspondents Association, Bangladesh (DCAB) at the EMK Center.
Her work focuses on the intersection of national security, AI, and election integrity.
She said cheap fake is a form of manipulated media where video, audio and images are altered using relatively simple and low-cost editing tools while deepfake is a form of synthetic media where video, audio and images are manipulated using artificial intelligence.
On cheap fakes, she said, in late 2020, China foreign ministry spokesperson Lijjan Zhao posted a fabricated image of an Australian soldier with a bloody knife next to a child.
"Days after the 2020 US presidential election, videos circulated on social media purporting to show election workers engaging in voting fraud. The misleading video circulated on twitter gathering attention from users and serving as doctored evidence that was fraud during the election. Local law enforcement investigated the location in the video to prove that it was false, " she said as examples of cheap fakes.
"AI generated images and videos are also used for satire and parody. Numerous deepfakes have circulated in the US presidential election which are clearly fake images used for humour," she said while giving examples of deepfakes.
There are tools to identity those deepfakes and cheap fakes.
The most sophisticated tools are in the domain of private companies and governments include sensity AI, Content Authenticity Initiative, Hugging Face, Deep Media, Deepfake-o-Meter, Reality Defender, and TrueMedia, she said.
Replying to a question on how the AI is being used in foreign policy practices, she said: "What I have noticed with the use of AI, AI works best if you have a problem or a challenge, you're trying to identify that AI can then help with.
"I would say technology works best as to leverage as a tool that's helpful when there's a problem or a challenge that it could serve as a solution for," she said.
"In terms of thinking of AI and national security within the US, the US government, particularly in the security area, have been using AI a lot longer than what we are aware of with ChatGPT's release in late 2022, mainly because they process a lot of data and it's not possible for an individual to go through that data.
"So instead of just having a software program, it makes it easier for them to bring various data points together to say, look for anomalies that may say that a terrorist attack is happening, for example.
"Or if you go to the Department of Homeland Security's website, they provide insight into the various ways that they're using AI within their law enforcement security operations, as well as within Department of Homeland Security is emergency response, so the Federal Emergency Management Agency, if disaster strikes, they respond to it, and so they're using AI within employees,” she said.
DCAB President Nurul Islam Hasib also spoke at the event.