AI Broke Trust: Why Identity Has to Step Up in 2026

In cybersecurity, we like to believe that trust can be protected with the right tools. Firewalls, endpoint protection, MFA, EDR—stack enough controls together and risk shrinks. But 2025 delivered a hard reality check: AI didn’t just improve attacks. It broke the trust models we depended on.

Phishing emails no longer look suspicious. Impersonation no longer sounds fake. Voice, video, and collaboration platforms now carry the same risk as a malicious attachment—sometimes more. Attackers didn’t bypass controls; they took advantage of the way people trust familiar requests and workflows.

In 2026, the most important security controls won’t live only at the login screen. They’ll show up inside calls, chats, approvals, and support requests—places we rarely treated as authentication problems before. Defending identity in 2026 means adding authentication where it never formally existed and dramatically strengthening it where it already does.

The Blind Spot: How We Authenticate Internally

For years, security programs have focused on authenticating users to systems. Organizations invested heavily in identity providers, MFA, conditional access, and device trust. But there is a critical blind spot in that model: we rarely authenticate people to each other inside the organization.

Most internal interactions still rely on informal trust signals—a familiar voice, a known name, or a message that appears inside Microsoft Teams or Slack. “Internal” is often treated as synonymous with “safe.” Attackers understand this and are deliberately targeting it.

Internal IT staff, help desk teams, managers, finance teams, and everyday employees are now impersonation targets. Once an attacker can convincingly pose as someone “inside,” even mature security programs can be undermined through routine workflows and social pressure.

Case Study: Retool—When Internal IT Can’t Be Authenticated

A single incident at Retool, a software company, ultimately led to the compromise of 27 customer environments and a reported $15 million loss at one affected customer.

What made this attack effective wasn’t malware or a zero-day exploit. It was trust in internal IT.

In this incident, attackers first interacted with an employee through a fake identity portal designed to look legitimate. But the most important part of the attack came next.

The attacker contacted the employee directly, claiming to be a member of Retool’s internal IT team. According to Retool’s incident response write-up, the voice was deep-faked to sound like a real employee and demonstrated familiarity with the office layout, coworkers, and internal processes. This wasn’t generic social engineering—it sounded like normal internal support.

Even as the employee grew uneasy, the interaction stayed within what felt like a valid internal workflow. Trust in internal IT—not a technical vulnerability—carried the attack forward. The attacker ultimately obtained an additional MFA approval, registered their own device, and maintained access.

What makes this incident so important for 2026 planning isn’t simply that MFA was involved. It’s that the organization had no strong way for employees to authenticate internal IT staff once the interaction moved into voice and conversation—exactly the blind spot attackers are now exploiting.

This Isn’t About Executives—It’s About Everyone

One of the most unsettling aspects of the Retool incident is who was impersonated. The attacker didn’t clone the voice of a CEO or public figure. They cloned the voice of a normal internal IT employee.

That fundamentally changes the threat model. Attackers are no longer limited to high-profile targets. They are impersonating everyday roles—IT support, coworkers, managers—anyone whose voice and behavior can be learned through meetings, calls, or collaboration tools.

If attackers can convincingly impersonate normal employees, familiarity and role can no longer be treated as security controls. Organizations must rethink how identity is verified between people, not just between users and systems.

Stronger MFA—and Mutual Authentication—in 2026

This doesn’t mean MFA is broken. It means how organizations implement and rely on it must evolve.

In 2026, stronger MFA means prioritizing:

• Phishing-resistant MFA, including FIDO2 security keys and hardware-backed authentication
• Passkeys, which eliminate shared secrets and dramatically reduce phishing and replay risk
• Biometric authentication, when used appropriately and backed by secure hardware
• Device binding and session controls that prevent attackers from silently registering new devices
• Step-up authentication for sensitive actions like access changes, financial approvals, and administrative tasks

Just as importantly, authentication has to become mutual. Today, organizations focus almost exclusively on authenticating users. But employees also need reliable ways to authenticate IT staff, service providers, and executives. Customers similarly need ways to authenticate vendors and partners, not just the other way around.

In the Retool incident and many similar IT impersonation scams, employees had no simple way to verify whether the person contacting them was truly internal IT. In 2026, that has to change. Practical approaches include requiring IT staff to reference a ticket number that employees can independently verify in a trusted internal system, or enforcing policies where sensitive support actions only occur through known platforms rather than ad-hoc calls or messages.

The goal isn’t to slow work to a crawl. It’s to give users clear, expected ways to verify identity when a request carries real risk.

Collaboration Tools Are the New Front Line

One of the most consistent attack patterns in 2025 involved collaboration platforms like Microsoft Teams. Incident response teams observed attackers:

• Flooding users with thousands of emails (“email bombing”)
• Anticipating IT tickets from overwhelmed users
• Contacting victims inside Teams while posing as help desk staff
• Requesting MFA approvals or remote access during the conversation

In many cases, Teams itself became the proof of legitimacy. It’s internal. It must be safe. That assumption is exactly what attackers rely on.

“Attackers are abusing the tools we trust the most,” said Matt Durrin during the podcast. “Teams, chat, internal systems—because users don’t question them the same way.”

Microsoft’s recent security feature updates reflect this shift in attacker behavior, as we discussed in LMG Security’s recent blog post, 5 New-ish Microsoft Security Features & What They Reveal About Today’s Threats.

Identity Takes Center Stage in 2026

Identity failures are no longer edge cases or theoretical risks—they are now one of the most effective ways attackers bypass even mature security programs. As organizations plan for 2026, identity needs to move to the center of security strategy—not just at the login screen, but embedded throughout internal interactions, approvals, and everyday workflows.

At LMG Security, we help organizations surface and test these identity assumptions through cybersecurity tabletop exercises and support long-term improvements through virtual CISO (vCISO) services focused on practical, risk-driven security programs. Let’s connect and discuss how we an help.

Identity has to step up in 2026. The question is whether your organization is ready.