Anthropic Red Team Simulates Self-Replicating AI Risks on Smart Appliances and Edge Devices
Anthropic's Red Team has published research simulating the threat of small AI models self-replicating across smart appliances and edge clusters, evaluating 26B