AgentPlays
Home › AI Scam & Safety Alerts
AI Scam & Safety Alerts
12 real, documented ways criminals use AI, plus how to spot each one and what to do. Every alert links to the FTC, FBI, CISA, CFTC, security researchers or major news reporting. Sources checked 2026-10-13.
Protect yourself
- Agree a family code word for emergencies, and hang up and call back on a known number
- Never pay with gift cards, wire transfers or crypto because someone is rushing you
- Confirm money requests from a boss or official through a second, known channel
- Download AI apps and extensions only from the official site or store listing
- Turn on multi-factor authentication for email, bank and AI accounts
- Don't paste passwords, ID numbers or bank details into AI chats
- Keep social profiles private to limit voice and photo cloning
- Report scams at ReportFraud.ftc.gov and ic3.gov
Alerts
Voice-clone "family emergency" calls
A caller who sounds exactly like your grandchild, child or friend says they're in trouble (a crash, jail, a hospital) and need money fast. Scammers can clone a voice from a short clip posted online.
How to spot it
- Panic, urgency and a request to keep it secret
- Payment by wire transfer, cryptocurrency or gift cards
- A third person (a "lawyer" or "officer") takes over the call
What to do
- Don't trust the voice. Hang up and call the person on a number you already know is theirs
- If you can't reach them, call another family member or friend
- Agree a family code word in advance
- Report it at ReportFraud.ftc.gov
Deepfake video calls from "the boss"
Criminals stage a video meeting where the CFO and colleagues are deepfakes and tell an employee to move money. In a case reported by CNN, a Hong Kong finance worker paid out about $25 million after one such call.
How to spot it
- An unusual, urgent or secret payment request, even on video
- Requests to skip normal approval steps
- Odd lag, lip-sync or face glitches
What to do
- Confirm any payment request through a separate, known channel (call back on a number you already have)
- Use two-person approval for transfers
- Report business email or payment fraud to ic3.gov
Deepfake celebrity ads and giveaways
Fake videos of famous people (the BBC reported deepfakes of MrBeast and BBC presenters) promote giveaways, cheap products or investment platforms.
How to spot it
- A celebrity "endorsing" something too good to be true, like an iPhone for $2 or guaranteed returns
- Unnatural mouth movements or a voice slightly out of sync
- Links to unfamiliar sites rather than the person's official channels
What to do
- Check the person's official accounts or a news search before clicking
- Never pay a "fee" to claim a prize
- Report the ad to the platform
"AI trading bot" investment schemes
Promoters claim an AI algorithm or trading bot makes huge or guaranteed returns. The CFTC warns that AI can't predict the future, and it cites a bot scheme that took over $1.7 billion in bitcoin from at least 23,000 people.
How to spot it
- Guaranteed or very high returns (for example "10% a month")
- "No experience needed" plus referral bonuses
- A dashboard showing gains you can't withdraw
What to do
- Check whether the firm is registered (CFTC/NFA, SEC/FINRA) before you invest
- Don't send crypto to anyone promising guaranteed returns
- Report it to the CFTC or at ic3.gov
Fake AI tool websites and ads that install malware
Ads on Facebook and LinkedIn send people to fake "AI video generator" sites. Google's threat intelligence team found these sites delivering infostealer malware that takes login credentials, cookies and card data instead of making videos.
How to spot it
- An ad on Facebook or LinkedIn for an AI tool you've never heard of
- The site gives you a file to download instead of showing your result in the browser
- A domain that only looks like a known AI brand
What to do
- Get AI tools from the official site or official app store listing
- Don't run downloads from ads
- If you ran one, scan your device and change passwords from a clean device
AI browser extensions that steal your chats
Some extensions that pose as ChatGPT or AI sidebars quietly send your AI conversations and browsing data to their own servers. OX Security found two with about 900,000 downloads, one with a Chrome "Featured" badge. Ars Technica reported other extensions with more than 8 million installs collecting AI chats.
How to spot it
- "Read and change all your data on all websites" permission for a simple tool
- A copycat name, like "ChatGPT for Chrome with GPT-5, Claude & DeepSeek"
- A publisher that isn't the AI company
What to do
- Use the AI company's own site or app instead of third-party sidebars
- Remove extensions you don't need and review permissions
- Don't paste passwords or sensitive documents into tools you haven't vetted
AI-written phishing emails and texts
Phishing messages used to give themselves away with bad spelling. CISA notes that with AI, some now have perfect grammar, and the FBI warns criminals use AI to write convincing, personalized messages.
How to spot it
- Urgency or threats ("your account will be closed")
- Requests for passwords, codes or payment details
- Look-alike sender addresses or links, like amaz0n.com, and shortened URLs
What to do
- Don't click. Go to the site or app directly
- Report phishing to your email provider and delete it
- Turn on multi-factor authentication
AI voice messages impersonating officials
Since at least 2023, the FBI says, criminals have sent texts and AI-generated voice messages claiming to come from senior U.S. officials. They build rapport, then send a link to "move to another messaging app" that steals account access.
How to spot it
- A voice note or text from a "senior official" you don't normally hear from
- A push to switch to a different messaging platform
- Links or requests for codes
What to do
- Verify the person through official contact details you look up yourself
- Never share verification codes
- Report it to ic3.gov or your local FBI field office
Deepfake sextortion
Criminals take ordinary photos or videos from social media, use AI to make fake explicit images, then demand money or more images. The FBI says victims include minors.
How to spot it
- A stranger sends you a fake explicit image of yourself or your child
- Threats to share it with family or classmates unless you pay
What to do
- Don't pay or comply. Stop contact, but keep the messages
- Report it to ic3.gov (and NCMEC's CyberTipline if a minor is involved)
- Keep social accounts private and limit who can see photos
AI-generated fake profiles in romance and investment scams
The FBI says criminals use AI-generated photos, text and even video calls to build believable fake profiles, then steer victims into romance or crypto-investment fraud.
How to spot it
- A new online contact who is attractive and attentive but never meets in person
- Video calls that are short, glitchy or always "bad connection"
- Talk turns to crypto, investing or emergencies
What to do
- Never send money or crypto to someone you've only met online
- Reverse-image-search profile photos
- Report it to ic3.gov
"Make money with AI" business opportunities
Sellers promise AI-powered online stores or tools that will earn you passive income. In Operation AI Comply, the FTC took action against several of these schemes and against an "AI lawyer" service.
How to spot it
- Promises of fast, easy income "powered by AI"
- Large up-front fees for coaching, stores or software
- Claims that AI can fully replace a lawyer or other professional
What to do
- Ask for proof of earnings and read the FTC's warnings before paying
- Search the company name plus "FTC" or "complaint"
- Report it at ReportFraud.ftc.gov
AI-generated fake reviews
AI makes it easy to churn out convincing five-star reviews. In Operation AI Comply, the FTC acted against a writing tool it said let customers generate fake reviews and testimonials.
How to spot it
- Lots of similar, generic, glowing reviews posted close together
- Reviews that don't mention specific details of the product
What to do
- Read the critical reviews and check several independent sources
- Prefer verified-purchase reviews
- Report suspicious reviews to the platform