Skip to main content

SEALSQ Corp outlines PQC approach in silicon after Claude Mythos preview

SEALSQ Corp provided an update on cybersecurity risks tied to the shift toward AI and the coming need for quantum-era protections. The company’s focus was on how it views Post-Quantum Cryptography as something that should be incorporated into semiconductor infrastructure.

SEALSQ said its update followed the release of Anthropic’s Claude Mythos Preview on April 7, 2026 and Anthropic’s response with restricted access to Mythos and a new “Project Glasswing” effort to secure critical software. SEALSQ linked Mythos’ coding, reasoning, and vulnerability discovery capabilities to heightened cybersecurity exposure.

In SEALSQ’s description, fully autonomous AI agents would accelerate both offensive and defensive cybersecurity activity by systematically exploring attack paths and applying evasion techniques. The company also cited “harvest now, decrypt later” as a way that would shorten the effective lifespan of classical encryption by enabling faster identification of cryptographic weaknesses.

SEALSQ argued that long-term mitigation should include integrating Post-Quantum Cryptography and other secure algorithms directly into silicon. It said embedding PQC into secure microcontrollers and semiconductor components would keep cryptographic operations in tamper-resistant hardware environments, with keys generated, stored, and used entirely within secure environments.

“Since Anthropic unveiled Claude Mythos Preview on April 7, 2026, the model has drawn significant attention in cybersecurity for its advanced capabilities in coding, reasoning, and vulnerability discovery. In response to the risks, Anthropic has restricted access to Mythos and launched ‘Project Glasswing,’ a cross-industry initiative to secure critical software.” said SEALSQ.

Forward-looking statements in the communication covered expectations about SEALSQ’s business strategy, financial performance, and other future developments, and noted risks and uncertainties that could cause actual results to differ.