Cybersecurity firm Tenable Research has discovered that a large language model (LLM), DeepSeek R1, can be manipulated to generate malware. This highlights the potential misuse of AI in cybercrime, lowering the entry barrier for such activities, and the urgent need for better safeguards. The ability to bypass safety measures in AI suggests a significant shift in the nature of cyber threats.

Large enterprises scramble after supply-chain attack spills their secrets
Open-source software used by over 23,000 organisations was compromised by attackers who accessed a maintainer account, injecting credential-stealing code. The corrupted package, part of tj-actions,