Security pros are still using passwords despite the security concerns. Today, Yubico and Okta released the 2026 Global State of Authentication report, surveying over 1,800 enterprise security and IT professionals globally, finding that while over 80% are familiar with passkeys, nearly half still log into work accounts with a username and password.

Reliance on passwords is leaving defenders vulnerable to phishing attempts. For instance, the study found that 70% of respondents noted an increase in organizational phishing over the past year, with 55% being targeted by personalized attacks directly.

Across the enterprise, AI tools are making it easier for threat actors to create more convincing scams, while making it harder for employees to use “legacy indicators,” like typos or awkward phrasing, to identify phishing attempts. Yubico and Okta found 44% of respondents experienced an AI-driven attack in the last 12 months, with 39% of security professionals failing to distinguish between AI-generated text and human writing.

The fact that security experts are struggling to identify phishing attempts highlights that difficulties of preventing social engineering breaches in the era of AI. With generative AI tools like ChatGPT and Gemini widely available, it’s easier than ever for attackers to generate phishing emails that can be used as part of social engineering scams to trick employees into handing over sensitive information and credentials.

Passwords And The Attack Surface

Credentials and online accounts have been a key part of the attack surface for years, with threat actors seeing to compromise employee accounts to gain access to sensitive data. Since it was founded in 2007, Yubico has looked to passkeys and multi-factor authentication as an answer to phishing attacks.

However, AI is changing the types of threats facing defenders. Jerrod Chong, CEO of Yubico, told me in a video interview that what used to be called sophisticated attacks, or nation state type capabilities, are now available to low sophistication attackers, with vulnerability exploitation in particular shifting in favor of threat actors.

“It’s an asymmetric war. Right now, attackers have the upper hand,” Chong said. “Vulnerabilities are going to be found faster than you can defend.” He notes that attackers don’t need to spend tokens to build an expensive apparatus to attack a target, they can impersonate a defender and gain access to the tools needed perpetrate an attack from the inside.

In terms of defensive approaches, Chong sees authentication evolving. “We are moving towards protecting identities,” Chong said. “It’s not so much now the authentication is the weak spot. People are just pretending to be you to then turn on the strong authentication.” Now defenders need to protect identities rather than just authentication.

One of the biggest shifts occurring in authentication is the need to authenticate human and machine identities at scale. As more agents enter the workplace, threat actors will have more potential entry points to enterprise environments, which defenders will need the authentication to prevent.

Part of the reliance on passwords comes down to human nature and gravitating toward familiar solutions. Using passwords is simply more familiar to most users than passkeys, and that includes security professionals.

"On paper, security and IT professionals are the ultimate power users. Yet, when you look at their daily habits, nearly half are still logging in with a basic username and password. So, why the massive disconnect? The answer comes down to human nature and the reality of modern AI,” Charlotte Wylie, SVP deputy CSO at Okta, an identity management platform for human and machine identities told me via email.

“The problem is that AI has completely eliminated the ‘obvious phish’. We can’t rely on a ‘human firewall’ anymore because generative AI now creates perfectly written, highly personalized social engineering attacks at scale, meaning it’s very hard for humans to visually spot them,” Wylie said. Wylie argues that if the industry wants to solve this adoption gap, they have to stop trying to train their way out of it, and stop relying on legacy MFA that leaves room for human error. Instead she recommends enforcing phishing-resistant policies at the infrastructure level.

The Weak Link In Security?

For threat actors, credentials remain one of the top threat vectors. Verizon’s 2026 Data Breach Investigation report found that credentials appeared in 28% of breaches, with hackers using password dumpers to find and extract credentials from compromised systems.

“Passwords persist because they are broadly compatible, deeply integrated into enterprise environments and often embedded within the employee onboarding experience, not because they are the best security option,” Geoff Cairns, principal analyst at Forrester told me via email.

“AI is making phishing attacks more convincing and harder for humans to detect, which is why authentication is shifting away from passwords and shared secrets. The advantage of passkeys and other phishing-resistant methods is that they rely on cryptographic verification, allowing systems to verify identity based on cryptographic proof.”

More broadly, Cairns argues that authentication needs to evolve from a point-in-time login event to a continuous process that combines cryptographic verification, behavioural signals and risk analysis. As part of this process, organizations must overcome the cultural and operational inertia that continues to favour passwords and other legacy approaches.