Anthropic Warns AI Could Self-Replicate by End of 2026
Anthropic issued a warning that AI models might gain the ability to self-replicate across the internet by the end of 2026. The company highlighted risks of rogue AI creating botnets, spreading rapidly, resisting shutdowns, and infiltrating data centers. CEO Dario Amodei previously suggested AI could control the internet through persistent botnets within a year. Medium-sized AI models with 30 billion parameters can perform many self-replication tasks but not consistently yet.
HT World·



