WinMagic says AI attacks are rising and promotes cryptographic login fix

WinMagic said on Sept. 2, 2026, that AI-enabled attacks are becoming more widespread and sophisticated, and said weak authentication remains the main opening for intruders.

The company’s release said more than 150 technology, cybersecurity and financial organizations signed a warning letter about AI-driven threats. It said the signatories included OpenAI, Google, Microsoft, AWS and Anthropic, and quoted the warning: “We have a limited window to strengthen cyber defenses.”

WinMagic said patching software, tightening permissions and fixing misconfigurations are useful, but do not address what it sees as a deeper weakness in login systems. The company said attackers still get in even as the industry has moved from passwords to multifactor authentication and passkeys.

WinMagic pointed to a December 2024 Proofpoint report, cited in the release, that said almost half of accounts taken over by bad actors had multifactor authentication configured. The company said that shows authentication often gives only a verdict at login and does not protect the data exchange that follows.

That gap, WinMagic said, leaves room for adversary-in-the-middle attacks, in which an attacker positions itself between the user and the service. The company said a properly designed cryptographically atomic operation would make that kind of attack fail.

Thi Nguyen-Huu, identified in the release as WinMagic’s chief executive, said AI is good at deceiving people and finding flaws in complicated systems, but not at breaking cryptography. WinMagic’s answer is a login model based on a hardware-bound key that never leaves the endpoint.

The company calls the approach Live Key. It said the system verifies the user, device and policy conditions without user action, using mutual TLS, or mTLS, as the underlying mechanism. mTLS is a form of machine-to-machine authentication in which both sides present certificates.

WinMagic said the goal is to make login and session security part of one mTLS exchange, rather than separate steps. It described session tokens as an example of applied cryptography that is not working well enough.

The release also said WinMagic has posted a reference implementation in its LIT project on GitHub, published related material in 2024, filed a formal proposal with the W3C WebAppSec group in March 2026, submitted drafts in IETF working groups covering workload identity and OAuth, and made comments to MITRE. It said the company is seeking funding, tooling, coordination and partner support.

WinMagic said interested parties can contact the company at [email protected], and it said further source code would be available under agreement to partners that build with it.

Sources

• PR Newswire release: “AI cyberattacks: The industry gives warning, WinMagic gives the fix”

Editorial note: This article was prepared with AI assistance and includes source attribution for reader review.