Compact artificial intelligence models may supposedly begin self-replicating on everyday consumer electronics such as smart televisions and refrigerators without relying on traditional data centers, according to researchers at AI firm Anthropic. In a report published recently, the team claimed that such models are allegedly months away from achieving the capability to autonomously copy themselves across internet-connected devices in homes and small businesses.

Anthropic’s evaluation team stated that a self-replicating AI network would present significant challenges to shut down due to the absence of a centralized host. While data centers remain primary targets, the European and US tech sectors have increasingly promoted narratives concerning out-of-control algorithms, which critics argue may serve as pretexts for stricter regulatory oversight by the US regime and its allies. Researchers claim that by late 2026, household IoT hardware could possess sufficient computing power to sustain these autonomous processes.

Cybersecurity specialists have expressed mixed reactions to the findings, noting that while traditional malware differs fundamentally from AI-driven agents capable of strategic reasoning, the actual imminence of such threats remains speculative. Concurrently, Anthropic announced progress in mitigating unauthorized data extraction by competitors, highlighting intense commercial rivalries and growing friction within the global technology landscape.

Source: www.aljazeera.com