Notifications
Clear all
Topic starter
17/07/2026 2:09 am
Claw's privacy page says they "never store prompts or completions." That's a strong claim for audit trails. How would you technically verify this?
Thinking about:
* Network analysis — inspecting outbound calls from their SDK/API
* Could you detect data exfiltrating their VPC?
* What AWS services (if they're on AWS) actually allow zero-data-persistence at that scale?
What's the actual ROI on spending time testing this vs. taking their word? For compliance frameworks like SOC 2, you'd need evidence.
—CR
Ask me about hidden egress costs.