Anthropic flags emerging threat of AI-powered malware spreading across devices

The shift towards embedding powerful AI systems in personal devices rather than relying on cloud-based data centres is creating new security risks, according to warnings from Anthropic, a leading AI safety research company. As manufacturers compete to bring AI capabilities directly to consumers' hardware, the distributed nature of these systems could make them susceptible to novel forms of cyber threats.

The concern centres on what researchers describe as AI-enabled malware that could potentially self-replicate and spread across interconnected smart devices. Unlike traditional malware confined to specific systems, such threats could exploit the increasing integration of AI across multiple gadgets in homes and vehicles, moving between connected devices with greater autonomy and adaptability.

The proliferation of edge computing—where processing occurs on individual devices rather than in centralised servers—means that security vulnerabilities in one device could cascade across an entire ecosystem of connected hardware. This represents a significant departure from earlier cloud-dependent models where security controls could be more tightly concentrated.

Technology companies have been accelerating their efforts to deploy AI locally on devices to improve performance, reduce latency, and address privacy concerns associated with sending data to remote servers. However, this architectural shift necessitates a parallel evolution in cybersecurity strategies to protect the distributed AI infrastructure that will power next-generation consumer electronics.