Anthropic's experimental AI, Claude Mythos Preview, has made headlines after identifying new mathematical vulnerabilities in cryptographic algorithms long studied by experts. However, cybersecurity specialists clarify that this does not mean the encryption securing online banking, messaging apps, or cryptocurrency wallets has been broken.

Significant Cryptographic Discoveries

The AI achieved two major breakthroughs:

  • HAWK Algorithm: Claude Mythos improved attacks on HAWK, a post-quantum digital signature candidate under review by the U.S. National Institute of Standards and Technology (NIST). It found a previously unknown mathematical symmetry, reducing the computational effort needed to crack keys and indicating that HAWK would require larger keys to remain secure.
  • Reduced AES Version: The AI developed a new technique called the "Möbius Bridge" that significantly enhanced attack speed against a seven-round version of AES-128. However, the full AES encryption used globally involves ten rounds and remains unbroken.

Implications for Internet Security

AES is the backbone of digital security for countless applications, including online banking, encrypted messaging platforms like WhatsApp and Signal, Wi-Fi networks, and government services. The discoveries by Claude Mythos do not compromise these systems today.

Modern security frameworks employ multiple layers beyond AES, such as authentication protocols and hardware security modules, ensuring robust protection for users.

AI's Growing Role in Cybersecurity Research

Remarkably, Claude Mythos initially resisted the challenge, deeming the task nearly impossible. After persistent guidance, it autonomously generated extensive outputs over three days, contributing original research in a field traditionally dominated by expert mathematicians.

Anthropic has also reported additional cryptographic attacks identified by Claude Mythos on other encryption algorithms, with ongoing verification.

What This Means for Kenyans

For everyday users in Kenya and worldwide, current encryption standards remain intact and secure. However, the findings underscore a future where AI accelerates both cybersecurity advancements and potential threats.

Experts advise continuing good security practices such as using strong passwords, enabling multi-factor authentication, and keeping software up to date.

Looking Ahead

The breakthroughs highlight how AI is reshaping cryptographic research, potentially enabling faster identification of vulnerabilities before malicious actors exploit them. This evolving landscape calls for vigilant updates to encryption standards and responsible management of powerful AI technologies.