The Salesforce MFA and Passkey Enforcement Rollout: What Admins Should Actually Be Doing Right Now

If your Salesforce org has been sending you security notification emails this year, you are not alone, and you are not imagining the confusion around dates. Salesforce has been rolling out a significant tightening of its multi factor authentication requirements throughout 2026, and the rollout itself has been paused, revised, and restarted more than once. For anyone responsible for keeping a Salesforce org compliant and accessible, it is worth separating what actually changed from what merely got delayed.

Two separate requirements, often confused with each other

There are effectively two distinct pieces of this enforcement, and conflating them is the most common source of confusion. The first is standard MFA for all employee users, requiring every non privileged internal user to complete multi factor authentication on every login. The second, and the one with real teeth, is phishing resistant MFA for privileged users, meaning anyone with the System Administrator profile or with permissions such as Modify All Data, View All Data, Customize Application, or Author Apex. According to guidance from an identity security vendor tracking the rollout, these two tiers were enforced simultaneously as of July 1, 2026, with standard MFA required on every login for all users and a materially higher bar, built on FIDO2 and WebAuthn standards, for anyone with elevated access.

The practical difference matters. Methods like SMS codes, email codes, time based one time password apps, and even Salesforce’s own push notification based Authenticator app are no longer sufficient for privileged users. What is accepted instead is a narrower set of phishing resistant options, specifically passkeys, built in device authenticators such as Touch ID, Face ID or Windows Hello, and hardware security keys such as a YubiKey. A practical breakdown from a Salesforce consulting partner lays out exactly which users this affects and what steps admins need to take in Identity Verification settings before the enforcement reaches their org, since users who have not registered a compliant method in advance are simply blocked at the login prompt when their turn in the staggered rollout arrives.

The pause, and why it happened

Here is where the timeline gets genuinely messy, and where the practical risk to organisations increases rather than decreases. On July 1, 2026, shortly after enforcement began, Salesforce hit pause on the standard MFA for all employee users piece specifically, after a defect surfaced in the phishing resistant enrolment flow. Users who already had a security key correctly registered were, in some cases, being incorrectly prompted to register an entirely new one, which risked locking out solo administrators and smaller teams at exactly the wrong moment.

A revised schedule followed almost immediately. As detailed by an independent analysis of the enforcement changes, the pause applied only to the broader employee wide MFA requirement, resetting the sandbox rollout to July 6, 2026 and the production rollout to July 20, 2026, with a longer staggered window than originally planned. Critically, the phishing resistant MFA requirement for privileged users and admins was never paused. It reached production on schedule on July 1, 2026, meaning any administrator without a compliant method registered was already at risk of being locked out during exactly this period of apparent breathing room.

Why organisations relying on Salesforce for regulated or sensitive data should not wait

Guidance aimed at nonprofit and mission driven organisations, which frequently manage donor, patient, or beneficiary data inside Salesforce, is unusually direct about the operational risk here. A Salesforce implementation consultancy’s assessment points out that an organisation with no compliant method registered for its privileged users when enforcement reaches their instance faces a genuine operational problem, not a minor inconvenience, since a locked out administrator during a live enforcement window can mean lost access to donor records, grant management data, or patient adjacent systems at a moment when the team can least afford it.

Beyond MFA itself, this enforcement wave is part of a broader tightening of Salesforce’s security posture that touches licensing and access decisions more widely. A review of related changes from a Salesforce security specialist firm notes that Salesforce’s own documentation defines phishing resistant methods as device bound, meaning a synced passkey stored in a password manager and shared across devices may not satisfy the requirement in the same way a hardware bound authenticator does. Organisations that use single sign on need to confirm their identity provider is passing the correct authentication signals, since Salesforce inspects those signals at login rather than simply trusting that SSO alone satisfies the requirement.

What licensing and governance teams should do this quarter

From a software asset management perspective, there are three concrete actions worth prioritising immediately, independent of exactly which date applies to your specific instance. First, identify every user in your org who holds the System Administrator profile or any of the four elevated permissions covered by the requirement, and confirm each one has at least two phishing resistant methods registered, so a single lost device does not create a lockout. Second, if your organisation authenticates through an external identity provider, confirm with your IT or identity team that the correct authentication method reference or context class reference signals are being passed through, since this is a common and easily missed gap. Third, build a documented lockout recovery process now, before you need it, covering exactly who can restore access to a locked administrator account and how.

It is also worth treating this enforcement wave as a prompt to review integration and service accounts, which often sit outside normal human MFA processes but frequently hold elevated permissions of their own. Any human being still logging in through a shared integration account, rather than their own named user, should be migrated off that account as a priority, since the exemptions that currently apply to non human integration users will not extend indefinitely to human logins routed through them.

Do not forget integration users and service accounts

It is also worth treating this enforcement wave as a prompt to review integration and service accounts, which often sit outside normal human MFA processes but frequently hold elevated permissions of their own. Any human being still logging in through a shared integration account, rather than their own named user, should be migrated off that account as a priority, since the exemptions that currently apply to non human integration users will not extend indefinitely to human logins routed through them. This is also a good moment to inventory every integration user in your org and confirm ownership, since these accounts are frequently created for a specific project years earlier, quietly retained, and then forgotten about entirely until a security review or, worse, an incident forces the question.

The audit trail question nobody asks until it matters

There is a licensing and compliance dimension to this rollout that is easy to overlook in the rush to simply get users enrolled. Once phishing resistant MFA is properly in place for your privileged users, that fact becomes something you can and should document as part of your organisation’s own security posture reporting, whether that reporting goes to an internal risk committee, an external auditor, or a client running their own vendor security assessment of your organisation. Being able to state plainly that your Salesforce administrators authenticate using hardware bound, phishing resistant methods, with evidence of enrolment dates and method types, is a meaningfully stronger answer than simply confirming that MFA is turned on in some general sense. If your organisation undergoes any form of SOC 2, ISO 27001, or client security questionnaire process, capturing this rollout properly now saves a scramble later when the next audit cycle asks for evidence.

There is also a cost dimension worth flagging for anyone tracking Salesforce spend closely. Hardware security keys are not expensive individually, typically running in the tens of dollars per device, but a team of any real size will need to budget for procurement, distribution, and a reasonable number of spares to cover lost or damaged keys, particularly for remote or distributed teams where shipping a replacement device takes days rather than minutes. This is a small, easily overlooked line item that is worth raising with whoever owns your IT hardware budget now, rather than discovering it only when the first admin loses a key during an active enforcement window and cannot log in until a replacement arrives.

How this compares with MFA enforcement elsewhere in your software estate

It is worth placing this rollout in the context of how other major enterprise vendors have approached similar mandates, because the comparison is genuinely useful for planning. Several large SaaS and cloud platforms have moved toward mandatory or phishing resistant MFA over the past few years, and the pattern across most of them has been broadly similar, an initial announcement with a generous runway, a staggered rollout by customer segment or region, and at least one round of delay or adjustment once real world enrolment friction surfaced. Salesforce’s approach here, including the mid rollout pause, sits comfortably within that pattern rather than outside it. The practical implication for a licensing and IT governance team managing multiple vendors simultaneously is that MFA enforcement announcements across your estate are rarely single events. They are more often the opening move in a rollout that will get revised at least once before it fully lands, and it is worth budgeting a small amount of planning slack into every such announcement rather than treating the first published date as final.

From a pure budgeting perspective, it is also worth flagging this rollout to whoever owns your annual IT security spend forecast, even though the direct licence cost impact is minimal. Phishing resistant MFA enforcement does not, on its own, change what your organisation pays Salesforce in subscription fees. It does create a small but real supporting cost in hardware procurement, helpdesk time during the enrolment window, and a modest amount of admin training, all of which are easy to overlook when the enforcement itself is framed purely as a security policy change rather than a project with its own resourcing footprint. Capturing that cost explicitly, even as a small line item, gives a more accurate picture of what compliance with this kind of platform level mandate actually costs your organisation each time a vendor introduces one.

Communicate the change clearly to the people it affects

A rollout with this many moving parts is also a good reminder that the technical fix and the communication around it are two separate jobs, and both need doing properly. Admins who receive a security enrolment prompt without prior warning tend to assume something has gone wrong, particularly given how much confusion this specific rollout has already generated across the wider Salesforce community on forums and social channels. A short, plain language notice sent to affected users before their instance reaches its staggered enforcement date, explaining exactly what they need to do, which methods qualify, and who to contact if they hit a problem, meaningfully reduces confusion and helpdesk load compared with letting each user discover the requirement cold at their next login.

The bottom line

The headline takeaway is straightforward even though the rollout schedule has been anything but. Phishing resistant MFA for Salesforce administrators and privileged users is real, it reached production on July 1, 2026 as planned, and it was not paused. What was paused, briefly, was the broader employee wide standard MFA requirement, and only until early revised dates in July. Treat the pause as a short delay to one specific tier of enforcement, not as evidence that the requirement itself is going away, and use whatever time remains before your org’s specific rollout date to get your privileged users properly enrolled.

More on the Blog