On July 30, 2024, a Michigan Medicine employee accepted an unsolicited multifactor authentication prompt, and a cyberattacker reached that employee’s email account and its contents. According to CBS Detroit, citing Michigan Medicine’s release, later review of job-related treatment traffic led to notices for about 57,891 patients. Coachable push-style MFA fails when a human can approve a fraudulent prompt once someone can start a login. Phishing-proof MFA 2.0 removes that approve surface: device-bound signatures leave nothing transferable to coach or fatigue. Closing the phishable login stops this path. Malware after a legitimate login is a harder, separate problem. For the full attack-chain narrative, read the companion on legacymfa.sucks.
FAQ
Would phishing-proof MFA have stopped the Michigan Medicine unsolicited prompt path?
Yes at the credential phase. The Michigan Medicine July 2024 email compromise followed the employee’s acceptance of an unsolicited multifactor authentication prompt. Phishing-proof, device-bound MFA has no coachable push or approval surface, so an unsolicited prompt cannot complete authentication for an attacker. Public reporting does not establish how the password or login context that triggered the prompt was obtained. Once a mailbox session is already issued, login MFA cannot undo mail already read; containment stays disablement, blocking, and revoke.
Why does approve-a-prompt MFA fail against unsolicited MFA push approval?
Approve-a-prompt MFA fails against unsolicited MFA push approval because the second factor is a human decision on a transferable challenge. Once an attacker can initiate sign-in, any approved prompt finishes authentication for them. Fatigue, habit, and social pressure all produce the same outcome: a legitimate-looking session on the attacker’s side. MFA 2.0 replaces that step with origin-bound public-key signatures that never leave the enrolled hardware. There is no prompt to spam and no code to type or read aloud. That is why legacy MFA was dropped for this failure mode.
What still needs controls after the push-approval path is closed?
After phishing-proof MFA blocks unsolicited push completion, residual risk is compromise on a machine that already holds a legitimate session. Fooling a user into approving a remote prompt is easy. Planting endpoint malware is not. At Michigan Medicine, once the attacker held mailbox access after the approved prompt, the organization disabled the account as soon as possible, blocked the attacker IP, and changed passwords. MFA does not recall messages already opened. Shorter email retention and stronger identity verification shrink blast radius; they do not replace removing the coachable factor.
How do passkeys compare to MFA 2.0 for this workforce email failure?
Passkeys are phishing-resistant at login: they bind the authentication ceremony to the real origin and remove typed OTP codes and coachable pushes at that step. MFA 2.0 is phishing-proof across the identity lifecycle, with no phishable factor at registration, device onboarding, authentication, or decommissioning. For the Michigan Medicine unsolicited-prompt failure, both remove the coachable push surface at authentication. The lifecycle gap matters when enrollment or recovery still uses email codes or helpdesk-issued secrets. Public reporting does not show enrollment abuse in this incident; the documented failure was the approve step itself.
What should replace coachable push MFA on workforce email after Michigan Medicine?
Workforce email should not rely on coachable push MFA after the Michigan Medicine July 2024 path. About 57,891 patients were later notified that health information in job-related email could have been exposed, per CBS Detroit’s summary of Michigan Medicine’s notice. Education and faster disablement help after the fact. The durable control is non-coachable, device-bound authentication that never presents a transferable approval. That is prevention, not detection: the unsolicited prompt path never completes.
