AI-Powered Threats: Deepfakes & AI Scams

The same AI that helps you defend yourself is being weaponized by attackers. Deepfake videos, cloned voices, AI-generated phishing, and entirely new attacks on AI systems themselves are reshaping the threat landscape. This lesson prepares you for the cutting edge — so you recognize the next generation of scams before they reach you.

What You'll Learn

• How attackers use AI: deepfakes, voice cloning, and scaled phishing
• How to detect AI-generated fakes and protect against them
• New attack types unique to AI systems, like prompt injection
• A verification habit that defeats even convincing AI fakes

Deepfakes and Voice Cloning

A deepfake is AI-generated or AI-altered video, image, or audio that convincingly impersonates a real person. Voice cloning needs only a few seconds of someone's speech — easily scraped from a social video — to produce a fake call in their voice.

The scams this enables are chilling:

• The "grandchild in trouble" call, now in your actual grandchild's cloned voice, begging for emergency money.
• The "CEO" video call instructing an employee to urgently wire funds — a real company lost millions to exactly this.
• Fake celebrity or executive endorsements promoting investment scams.

How to protect yourself:

• Agree on a family "safe word" that a real loved one would know but a scammer never could.
• Verify through a second channel. Hang up and call the person back on their known number.
• Watch for urgency plus money — the universal scam signature, deepfake or not.
• Look for visual glitches in video: odd blinking, mismatched lighting, blurry edges around the face — though these clues are vanishing as the technology improves.

Ask Perplexity to stay current:

What are the latest types of deepfake and voice-cloning scams targeting ordinary people, and what do experts recommend to protect against them? Cite recent sources.

AI-Scaled Phishing and Fraud

Beyond deepfakes, AI lets attackers operate at industrial scale. They can generate thousands of personalized phishing emails, build convincing fake websites in minutes, run chatbots that impersonate customer support, and create fake but realistic product reviews and social profiles.

The defensive mindset shift, which you already learned for phishing, matters even more here: judge actions, not appearances. Polished writing, a professional-looking site, and a confident voice are no longer evidence of legitimacy, because AI manufactures all three cheaply. Trust is earned through independent verification, not through how convincing something looks.

New Attacks on AI Itself

As you and the world adopt AI tools, attackers target the AI systems too. You should recognize these terms:

• Prompt injection — hiding malicious instructions inside content an AI reads (a web page, a document, an email) so the AI is tricked into ignoring its rules or leaking data. If you have an AI assistant that browses or reads your files, be cautious about what you let it process.
• Data poisoning — corrupting the data an AI learns from so it behaves badly.
• Jailbreaking — crafting prompts that bypass an AI's safety guardrails.
• Sensitive data leakage — pasting confidential information into a public AI, where it may be stored or surfaced later. (This is why the course keeps repeating: redact before you paste.)

You do not need to defend against all of these as a beginner, but understanding them makes you a far smarter, safer AI user — and these are exactly the topics employers are hiring people to address.

The One Habit That Beats AI Fakes

Technology to detect deepfakes will always lag behind technology to create them. So your most durable defense is not a detector — it is a verification habit:

• For any urgent request involving money, access, or sensitive data, stop and verify through an independent, trusted channel before acting.
• Treat unexpected calls, videos, and messages as claims to be verified, not facts.
• Build the family safe word, the call-back rule, and the "I will confirm and get back to you" pause into your reflexes.

This single habit defeats deepfakes, voice clones, AI phishing, and old-fashioned scams alike — because they all depend on you acting on trust without verifying.

A Quick Hands-On Exercise

Run this in Claude or ChatGPT to build your personal defense plan:

Act as a security coach. Explain the top 3 AI-powered scams an ordinary person might face this year, and give me a simple, memorable defense for each. Then help me design a 'family safe word' system and a verification rule I can teach my parents and friends.

Then actually share the safe word and call-back rule with your family this week. Defenses only work if the people around you know them too.

Your Homework for This Lesson

Set up a verification protocol with at least one family member or close friend: a safe word and an agreement to always call back on a known number for any urgent money request. Then ask Perplexity for the most recent deepfake scam trends so your knowledge stays current. You are now defending against threats most people have not even heard of yet.

Key Takeaways

• Deepfakes and voice cloning let attackers convincingly impersonate loved ones, executives, and celebrities.
• AI scales phishing and fraud, so judge requests by what they ask, not by how polished they look.
• New AI-specific attacks include prompt injection, data poisoning, jailbreaking, and sensitive data leakage.
• Detection tech lags creation tech, so a verification habit — stop and confirm via a trusted channel — is your best defense.
• Set up a family safe word and a call-back rule, and keep current with sourced tools like Perplexity.