On May 13, 2026, Palo Alto Networks published an advisory for CVE-2026-0257, an authentication bypass in the GlobalProtect VPN portal and gateway that lets an attacker “bypass security restrictions and establish an unauthorized VPN connection.” Fixes were available that month. In June, the company’s Unit 42 team observed active exploitation, and its October 20 update names the Qilin and Inc ransomware groups among those using it: bypass MFA to get a VPN session, move over RDP and SMB, take the data, deploy the ransomware. Everything after the first step rested on one assumption: a session already inside could move freely. Removing that assumption is what zero trust for a 50-person firm means.

What zero trust actually means

The definition comes from NIST Special Publication 800-207, published in August 2020: a collection of concepts for enforcing least-privilege, per-request access decisions in a network viewed as already compromised. The same document says zero trust “is not a single architecture but a set of guiding principles for workflow, system design and operations,” and that it cannot be achieved by replacing technology wholesale. I’d ask every owner to read that sentence before the next vendor meeting; nobody sells zero trust in a box.

In plain terms, the network stops vouching for anyone. Every request is checked against who is asking, what device they’re on and what they’re allowed to reach, and the check repeats rather than happening once at login. Three of NIST’s seven tenets do most of the work: location never implies trust, access is per session with least privilege, and no asset is trusted by default.

The five things a 50-person firm stops trusting

CISA’s Zero Trust Maturity Model, version 2.0 from April 2023, was written for federal agencies, and CISA says all organizations should consider it. Its five pillars name what a company of fifty trusts without having decided to.

Identity. You stop trusting a password. Every account gets MFA, the VPN and every administrator included, and shared logins go away. Microsoft found in 2019 that MFA can block over 99.9 percent of account compromise attacks, and CISA’s October 2022 guidance puts phishing-resistant methods such as FIDO keys first for administrators. The same rules cover service accounts and AI agents holding credentials nobody reviews.

Devices. You stop trusting a laptop because the company paid for it. Devices get enrolled, encrypted and patched, and only compliant ones open email and files.

Networks. You stop trusting the inside. A flat network where any workstation reaches the servers and backups is what let the GlobalProtect intrusions spread over RDP and SMB. Segment it, and move from an always-on VPN to per-application access where possible.

Applications and workloads. You stop trusting an application because it sits behind the firewall. Each system goes behind single sign-on with the same identity and device rules, and standing administrator accounts are removed.

Data. You stop trusting that sensitive data stays where it was filed. Know where client records, payroll and contracts live, label them, and restrict what can move them, AI tools included.

What it costs

For a 50-person firm on Microsoft 365, the first stage costs nothing in licenses. Microsoft’s security defaults come at no extra cost and require every user to register for MFA, require administrators to use it and block legacy authentication. The next stage, Conditional Access, which ties access to conditions such as the device and the location, requires Microsoft Entra ID P1, listed at $7 per user per month on an annual commitment and also available to Microsoft 365 Business Premium customers. At list price, P1 for 50 people is $4,200 a year.

The real cost is decisions. Which legacy application can’t do modern authentication, and whether you replace it or wall it off. Which people hold domain administrator rights all day. Whether the VPN appliance stays as the front door at all. Each is a conversation, not a line item, and each stalls when nobody at the top owns it. Against that, Verizon’s 2026 Data Breach Investigations Report found ransomware in 48 percent of breaches, with a median payment of $139,875 among victims who paid, and notes that small organizations are disproportionately affected. IBM’s 2026 Cost of a Data Breach Report puts the global average breach at $4.99 million, a record.

Where to start

NIST’s implementation guide, SP 1800-35, published in June 2025, documents 19 example builds made from off-the-shelf products. None did everything at once. The evidence points to this order:

  1. MFA on every account, phishing-resistant for administrators first. Credential abuse still opened 13 percent of breaches in the 2026 DBIR.
  2. Patch the edge. Exploitation of vulnerabilities is now the most common first step in a breach at 31 percent, and only 26 percent of critical known-exploited vulnerabilities were fully fixed in 2025.
  3. Inventory and enroll every device, then require compliance before access.
  4. Remove standing administrator rights and shared accounts; give every exception an owner and an end date.
  5. Segment the network so a compromised workstation can’t reach backups, finance systems or servers.
  6. Collect the logs that would show the next intrusion, and have someone read them. If a provider runs your network, these six items belong on the evidence list you ask them for.

Frequently asked questions

Is zero trust a product we can buy?

No. NIST SP 800-207 calls zero trust a set of guiding principles rather than a single architecture, and says it can’t be achieved by replacing technology wholesale. Products enforce the principles; the decisions are yours.

How much does zero trust cost a small business?

For a firm on Microsoft 365, security defaults cost nothing extra, and Conditional Access requires Microsoft Entra ID P1, listed at $7 per user per month on an annual commitment, or a Business Premium license. The larger costs are time and the decisions about legacy applications and administrator rights.

Do we still need a VPN under zero trust?

Sometimes, for specific systems, but it is no longer the trust boundary. CVE-2026-0257 showed what happens when a VPN session stands in for identity: attackers bypassed MFA and moved across the network. Under zero trust, each application checks the user and device itself, so a VPN session alone opens nothing.

Zero trust for a 50-person firm comes down to one sentence: the network no longer vouches for anyone, so every door checks for itself. The licenses are mostly already paid for. What remains are five decisions, an owner for the exceptions list, and the discipline to patch the edge before the attackers arrive. That is the whole program, and it fits a company of fifty.