Voice and video deepfakes
A call or video where the voice and face are an exact match — but the person on the line is an AI clone trained on a few seconds of public audio.
Pre-agreed proof of identity. Free.
Verify the people you trust — even when their voice, account, or face has been faked .
Stolen accounts, cloned voices, deepfaked video, and social engineering can be stacked into a single attack — the attacker has your wire, your voice, your relationship context. The only proof that survives is a word you agreed offline, that was never on any channel they could compromise. This site helps you create one, agree on when to use it, and practice together.
- Free, forever
- No account, no email
- Runs entirely in your browser
- For families, teams, friends, sources — anyone who matters
The threat, in numbers
This isn't a hypothetical. It's already happening.
$2.7B
Lost to impersonation scams in the United States in 2023.
FBI Internet Crime Report
3 seconds
Of audio is now enough to clone a person's voice convincingly.
Microsoft VALL-E research
700%
Increase in reported deepfake incidents since 2023.
World Economic Forum
What it defends against
Account compromise + impersonation = unbeatable evidence. Unless there's a word the attacker never had access to.
A safeword isn't just a defense against AI voice clones. It's the one identity check that survives every modern attack — because the proof was agreed offline and was never on any wire to intercept.
The killer combination
An attacker with a stolen account has your channel. With AI voice cloning, they have your voice. With access to your messages, they have your relationship context. Stack those and the call, the email, or the text is indistinguishable from the real you. The only thing left that they can't fake is a word the two of you agreed on, in person, that lives nowhere a system can leak.
Account takeover
Email, Slack, SIM, or social — a hijacked account sends a message from the real address to the right people, with the right history. The channel checks out. The person doesn't.
Pretexting
“This is IT.” “This is your bank.” “This is the CEO.” A confident voice, an urgent ask, the right names. The script is the whole attack.
Vendor and invoice fraud
A real expected invoice arrives at the right time from the real vendor's email — but the wire instructions have been silently switched.
Personal safety check-ins
First dates, travel, anyone going somewhere where “I'm fine” needs to be more than two words. A pre-agreed phrase signals safe — or signals duress.
Source and journalist verification
When a contact reaches out on a new number or platform, a pre-agreed word is what proves it's still them — not someone who's seized their device or burned their identity.
How it works
Three steps. About five minutes. Then you're protected.
No app to install. No account to manage. Do it once with the people you trust, and you're done.
Step 01
Generate a safeword
Our tool creates a memorable, phone-safe verification phrase. It's designed to be easy to recall under stress, and clear on a bad phone line.
Open the generator →About 1 minute
Step 02
Agree on a protocol
Decide together when the safeword must be used — wire instructions, account changes, urgent asks, unfamiliar channels, anything that feels off. Works the same for a household, a finance team, or a small group of friends.
Build a protocol →About 2 minutes
Step 03
Practice together
Run short, realistic drills so each person knows exactly what to do when a suspicious call comes. Practice turns instinct into muscle memory.
Start a drill →About 2 minutes
Recommended by
The FBI, FTC, and AARP all recommend establishing a secret verification word with the people you trust — for families, finance teams, and any small trust group where impersonation is on the table.
FBI — Internet Crime Complaint Center FTC — Consumer Alerts AARP — Fraud Watch Network
Bronze Urchin 40
Try it now
A live tool. Nothing is sent to our servers.
Every word is generated in your browser using a curated list chosen for clarity over the phone. Close this tab and nothing remains — only the people you share it with will ever know.
- MemorableShort, unusual words that stick even under stress.
- Phone-safeNo homophones or sounds that garble on bad audio.
- StrongEnough entropy that guessing it in a single call is effectively impossible.
- PrivateGenerated client-side. Nothing leaves your device.
Real stories
What happens when people don't have one.
These accounts are drawn from public scam reports. Every one of them would have ended differently with a safeword.
My mom got a call from “me” crying, saying I'd been in a car accident. She almost wired $5,000 before my dad thought to call me directly.FamilyReddit, 2024
The voice sounded exactly like my grandson. Every inflection, every pause. If we hadn't had a code word, I would have given them everything.Grandparent scamAARP Fraud Watch
Our CFO's voice was cloned from a conference recording. If we hadn't had a verification word, we'd have wired $200,000 to a scammer's account.BusinessFinance director, mid-size tech company
My friend and I have a check-in word for first dates. When someone claiming to be her called me, they couldn't give it.Personal safetyReddit, 2025
Common questions
What you probably want to know.
Does Safewords store or see my safeword?
No. The generator runs entirely in your browser. We don't have servers that process your word, we don't log it, and we don't use analytics that would capture it. When you close the tab, there's no trace on our end.
Who should I share a safeword with?
Anyone who might one day receive an urgent, out-of-the-ordinary call from you or about you — immediate family, close friends, and for organizations, anyone with authority to move money or grant access. Many families set up one safeword for the whole household.
What if someone overhears the safeword?
Change it. A safeword is only useful while it's secret. We recommend agreeing on a way to rotate it — in person, in writing, or through another channel you trust — if you suspect it has been compromised.
Can I use this for my company?
Yes. Many businesses use shared verification phrases alongside other controls to prevent wire fraud, vendor impersonation, and social engineering. The same principles apply: generate, agree on the protocol, and drill.
Is a safeword really enough?
It's one layer, and it's a very effective one. It should sit alongside other good habits — never rushing a large transfer, calling back on a known number, and being skeptical of urgency. A safeword is the fastest way to confirm identity when a voice, video, or account alone is no longer trustworthy.
Is this just for families?
No. Anything that can be used to ask family members for money, gift cards, or sensitive information can also be used on a finance team to authorize a wire, on an executive to phish credentials, on a journalist to verify a source, or on anyone in a small trust group. The defense is the same: a pre-agreed word the attacker never had access to.
Is this only about AI voice cloning?
No. Voice cloning is the loudest reason right now, but the same defense covers account takeover (a hijacked email or Slack message that looks correct), deepfaked video calls, pretexting (“this is your bank”), vendor invoice fraud (real vendor, swapped wire details), and SIM-swap impersonation. Any attack where the channel itself can be faked is solved by a verification step the channel can't carry.
Protect the people you trust. Today.
It's free. It takes about five minutes. Nothing is stored, and no account is required.