Curated news, CVE analysis, and threat reports from the world's top cybersecurity sources.
Researchers propose that achieving stronger AI safety requires moving beyond treating AI models as inscrutable black boxes and instead identifying specific cognitive elements and reasoning structures within them. By mapping internal representations — such as what the model attends to, how it chains reasoning steps, and where its knowledge of concepts like deception or harm reside — developers could gain interpretability necessary to verify safety properties before deployment.
Anthropic's Claude Mythos Preview derived an end-to-end key-recovery attack on HAWK-256 (expected runtime ~3h42m on 96-core server) exploiting a previously unused symmetry in the lattice scheme, and achieved a 200-800x speedup for a 7-round AES-128 attack by removing a 256-way guessing step. Neither result affects production systems — HAWK remains a NIST post-quantum candidate and the AES attack targets only 7 of 10 rounds requiring impractical chosen plaintexts.
Anthropic’s Claude Mythos Preview AI has successfully cracked cryptographic schemes, including a post-quantum test candidate and a reduced-round AES-128 attack. The AI derived a key-recovery attack against HAWK-256, a lattice-based signature scheme in NIST’s post-quantum standardization process, reducing the expected work factor from 2^64 to 2^38.
Amazon Web Services (AWS) offers two key cryptographic key management solutions: AWS Key Management Service (AWS KMS) and AWS CloudHSM. AWS KMS is a fully managed service designed for most cloud-based key management needs, integrating seamlessly with all AWS services and regions. It is cost-effective, charging $1 per key plus $0.03 per 10,000 requests per month, and supports symmetric and RSA 2048 operations.
This article compares AWS KMS and AWS CloudHSM for cryptographic key management. AWS KMS is a fully managed service suitable for most cloud-based key management needs, while AWS CloudHSM provides dedicated HSM instances for legacy applications and strict compliance requirements.