This week the story small teams kept running into was not a flashy breach write-up. It was the quiet kind: someone calls the help desk, sounds exactly like a principal, a finance lead, or a vendor contact, and asks for a password reset, an MFA change, or a quick exception. AI voice cloning has made that pressure call cheap and convincing. Schools and 40-person shops feel it first because a single overloaded inbox or after-hours phone line is often the whole gate.
The pattern is simple. The caller already knows a name, a ticket number, or a recent outage. They push urgency. Staff want to be helpful. If your reset process still treats a familiar voice as proof, the attacker does not need malware. They need five minutes and a cloned clip. You do not need a perfect attribution report to act. You need a rule that voice alone never moves access.
A familiar voice is not a credential.
What to do on Monday
Pick the one path attackers actually use against you: phone or chat requests that change login power. Write the stop-rule in plain language, tell everyone who answers the phone, and run one short drill before lunch. No new platform required.
- Post a one-line policy at every help-desk seat: no password resets, MFA changes, or inbox forwarding from voice or chat alone.
- Require a second channel you already own (manager approval in email or a known internal ticket) before any account recovery completes.
- Build a five-question call sheet: full legal name, employee or student ID, ticket number, callback number on file, and a shared secret that is not public.
- Run a 10-minute tabletop with whoever answers phones: one fake urgent call, practice the refusal, document where the process broke.
- Turn on or confirm logging for admin password resets and MFA device changes so you can see odd bursts the same day.
Containment here is boring on purpose. If Monday ends with a written voice-is-not-proof rule and one practiced refusal, your office is harder to con than it was on Friday. That is the whole job this week.