Anthropic has implemented preliminary egress bandwidth controls as part of ASL-3 security, exploiting the large size of model weights so that rate limits on outbound network traffic can detect and block weight exfiltration even if an attacker has otherwise significantly compromised internal systems.

supports (1)

• SActivating AI Safety Level 3 protections primary_testimony
  weight

  0.90

  locator: Security section

  “By limiting the rate of outbound network traffic, these controls can leverage model weight size to create a security advantage... we expect to get to the point where rate limits are low enough that exfiltrating model weights before being detected is very difficult—even if an attacker has otherwise significantly compromised our systems.”