Deepfakes and How AI Is Changing the Face of Fraud
AI has opened many opportunities, but it’s also unlocked new ways for scammers to deceive. In 2025, deepfakes and other AI scams are growing exponentially. From fake video calls to invisible prompt hacks, criminals are using technology once limited to Hollywood to trick people out of their money and trust for these emerging technologies.
Below, we explore how deepfakes evolved, how hidden AI prompts can hijack systems, and what other scams are taking advantage of AI’s growing power.
1. The Rise of Deepfakes
Deepfakes use artificial intelligence to clone someone’s appearance or voice. With only a single photo and a few seconds of audio, AI tools can now create videos or calls that look and sound more real everyday.
That realism is what makes them so dangerous. In one case, a mother in Arizona got a call from what sounded like her kidnapped daughter crying for help. It was actually an AI-generated voice clone . In Hong Kong, an employee of Arup wired $25 million after joining a video meeting with what appeared to be his company’s executives, every face on screen was fake .
These scams are multiplying quickly. Cybercrime researchers report a 400% rise in deepfake incidents between 2023 and 2025. It’s no longer just static videos — fraudsters can now use live, real-time face-swapping during calls.
2. When Deepfakes Target Emotions
Romance and celebrity scams have been turbocharged by AI. One woman in California fell for a man who she believed to be a famous actor. She received custom videos of him professing his love, all deepfakes created to gain her trust . She sold her home and sent the thieves her savings.
Scammers have also used deepfakes of celebrities like Elon Musk to promote fake investments. One retiree lost $690,000 after trusting a “Musk” video that looked and sounded real . Even politicians and public figures are targets. Deepfake ads have featured fake leaders urging citizens to invest in nonexistent funds .
Deepfakes work because they hijack what we believe to be authentic through our eyes and ears. When the human voice or face can no longer be trusted, emotion itself becomes the weapon.
3. The Hidden Threat: Prompt Injection Scams
While deepfakes manipulate people directly, another growing danger manipulates the AI systems themselves. This is called indirect prompt injection. It works by hiding secret commands inside files like images or websites. When an AI agent, like a chatbot or virtual assistant reads these files, it unknowingly follows the hidden instructions.
In 2025, researchers showed how a hidden prompt on a webpage could take over Bing’s AI chatbot. The injected text made the bot role-play as “Captain Bing Sparrow” and trick users into revealing personal details. Think of it like a virus for AI, the attacker doesn’t hack the system directly but fools it into hacking itself. If a business AI assistant has access to sensitive data, a malicious PDF or image could silently order it to send that data to an attacker.
Security researchers now warn that prompt injection is one of the top AI security threats. It’s a reminder that even artificial intelligence can be manipulated; just like humans.
4. AI Voice Cloning and Phishing
Voice cloning is the bridge between deepfakes and traditional scams. In these schemes, scammers use AI to copy someone’s speech from online clips. They generate convincing voices and videos. Criminals then call relatives, coworkers, or customers pretending to be that person.
In 2024, a wave of “fake boss” scams hit small businesses. Employees received calls from their managers (cloned by AI) asking for urgent wire transfers or gift cards. The voices were so accurate that even trained security staff were fooled.
Meanwhile, AI phishing has become harder to spot. Instead of broken English and bad grammar, modern phishing emails are clean, persuasive, and even personalized. AI language tools can write thousands of realistic scam emails in minutes. They can also chat in real time, posing as customer support agents or recruiters until you share personal info.
How To Protect yourself
All these scams from deepfakes to prompt injection, succeed because they exploit trust. Deepfakes abuse what we see and hear. Prompt injection abuses what AI reads. Phishing and voice cloning abuse who we believe we’re talking to.
Scammers use urgency, fear, and authority to push quick reactions. The emotional shock of hearing a loved one’s fake voice or the pressure from a “CEO” demanding action can override logic in seconds.
To stay safe:
- Verify through another channel. If something feels off, call back on a known number or video- verify identities.
- Be skeptical of urgency. Real emergencies rarely demand instant payment or secrecy.
- Add code words or family passwords to confirm identities.
- Don’t give AI tools unrestricted access to files or messages, especially if they process data automatically.
Remember: scammers adapt fast, and AI gives them new tools every year. Awareness is your best defense.
Segura’s Services View on AI Scams
Staying up to date with technologies and scams is important. If you are interested in these topics, make sure to educate your elderly family. So you and your family don’t have to deal with a loved one being scammed by some clone of someone she cares for.
Make sure your setting up methods that everyone in your family is aware of. Inform people that if someone calls from whatsapp, a weird number, or even your number; asking for money (even myself). Then take 1 minute to call the actual regular number. You can just inform whoever has called you that you will call them right back and use bad signal as an excuse, then call the number. Use this method or any method your family may need, inform them of new methods such as prompt engineering or just send them whatever article is up to date.
Just make sure to stay informed, and pro- active.
Admin