A few years ago, if you saw a video of someone saying something, you believed it. That's just how video worked. Today, that assumption is quietly falling apart, and most people haven't caught up to it yet.
This technology is known as deep fake. Video, audio or images created with AI that make it look as though a real person said/said or did/did something that they did not. The figures behind it, fortunately, are a little frightening.
 |
| Credit: https://pixabay.com/ |
So, What is the Definition of a Deepfake?
The results of a deepfake are generated using a supervised method of AI that analyzes hundreds or thousands of images, video snippets or audio snippets of a real individual, and mastering a way to convincingly mimic their face, voice and mannerisms. The name is derived from the union of "deep learning" (one of the AI techniques) and "fake.
Previous iterations of the technology were cumbersome, circa 2017-2019. There was typically a lack of congruence; a blinking that didn't seem to, a lighting that was incongruent, a voice that seemed robotic. This doesn't hold true anymore. Today, voice-cloning technologies require only a few seconds of a person's voice to create an accurate replica. The quality of the videos has also improved as rapidly as the quality of the videos has improved.
The Numbers Are Bigger Than Most People Realize
This is no Internet curiosity, because:
- Today, an estimated 8 million deepfakes are circulating online, which is about 16 times as many as were estimated 2 years ago.
- In North America, deepfake scams have increased by 3,000% in one year and account for a significant proportion of all fraud attempts worldwide.
- One of the biggest isolated financial damages of this sort of scam occurred when a finance employee of a company was invited to what appeared to be a regular video call with a few senior colleagues, which were all found to be deepfakes. The company got cheated out of more than $25 million.
- Perhaps most alarming: research indicates that normal people can only identify a well-constructed deep fake approximately 0.1% of the time. That is, rarely, if ever.
- The research indicates that you would not know it in your gut, which is why it matters.
 |
| Credit: https://pixabay.com/ |
It's not only about scams, there's a darker side.
Financial fraud is all the rage right now, but the first time and most common use of deepfake technology was not financial, it was non-consensual explicit content, typically of a female target, created without their knowledge or consent. Initial research into online deepfakes revealed that almost all were of that type. This is still a big problem today, and it's the kind of damage that led a number of countries to enact emergency laws.
It's important to be aware of this because “deepfakes” is often interpreted as a tech-fraud story, but the damage to the victim's reputation, harassment and psychological effects have been equally prevalent, if not more frequent.
Why This is Getting Worse, Not Better?
There are three things that are all coming together:
1. Tools becoming more user friendly. Coding skills or costly software are no longer necessary. There are apps that will allow anyone to produce a convincing fake video or voice recording in just minutes, sometimes for free.
2. Detection is lagging behind. Developed to detect deepfakes, AI technologies can achieve high accuracy rates, reaching above 90% in laboratory testing. However, in the real world facing new and changing techniques, that precision can drop significantly and sometimes by almost 50 per cent. The battle between detectors and fakes is an ongoing one: the more sophisticated the detectors are, the more sophisticated the fakes are.
3. Most people (businesses) are not ready. Many companies still don't have a strategy in place for dealing with a deep fake video call or voice message that demands payment or confidential information from an employee. This is precisely what scammers are taking advantage of.
Also read: Is studying still worth it in the age of AI?
In some ways, governments are finally catching up.
In some respects, governments are finally catching up.
2025 and 2026 are years of turning points in regulation. The European Union's AI Act, which mandates the easy identification of AI-generated content, will impose potentially tens of millions of euros in penalties for failures. Since 2022, dozens of states in the United States have enacted their own laws addressing deepfakes, with some statutes explicitly addressing non-consensual explicit deepfakes and providing a quicker remedy for the victim to seek removal of such content.
This is a good trend, but enforcement efforts are only now beginning to respond to the magnitude of this issue. It takes time for laws to be carried out internationally, and much of this material is generated and disseminated quicker than platforms or courts can catch up.
 |
| Credit: https://pixabay.com/ |
The simple steps you can take to protect yourself.
It is not necessary to be constantly aware of all the videos you watch, but there are a few habits that actually help:
- Be especially wary of pressing financial demands from someone who says it's a person you know and who looks and sounds like that person. The trick in getting people to wire money is to make them feel the urgency to do so, bypassing the rational thought that suggests a second opinion. If the message seems like it was sent quickly, dial the person's saved number instead of the number provided within the message.
- Have a verbal "safe word" with close family members or coworkers who deal with money that a scammer would not know in case of emergency.
- Be wary of an online video call that appears legitimate, particularly in a business context where you are doing a financial deal. If it doesn't feel quite right, check it out first.
- If you or someone you know is a victim of content that you do not want to see, report it immediately. There are more places than ever where platforms are legally obliged to respond promptly to requests for removal. Platforms now face more obligations than ever to respond quickly to requests for removal.
Conclusively, what's happening with deepfakes is part of a broader trend with AI: the technology for making convincing fake content is advancing faster than our collective ability to detect it and even faster than the laws enacted to address it. It is not as if video and audio can't still help prove the truth, but the old saying “seeing is believing” is not applicable anymore.
إرسال تعليق