IGNITE9 LLC
What to ask about device-local AI and private messaging
Evaluate where inference happens, what leaves the device, and how encryption interacts with AI features.
IGNITE9 EDITORIAL · PRIVATE INTELLIGENCE
Map the actual data path
“On device” describes a location for computation, not an entire privacy policy. Ask whether prompts, outputs, telemetry, crash reports, backups, or optional cloud features leave the device. Request a clear description of each processing path.
Confirm device coverage
Local models depend on memory, processor capabilities, battery use, and model size. Ask which phones and operating systems are supported and what happens when a device cannot run a feature. A fallback service may have different privacy implications.
Understand the encryption boundary
End-to-end encryption protects a communication path, while an AI feature may operate before encryption or after decryption on the device. Clarify who can access plaintext, how group membership works, and whether AI features change the threat model.
Review claims product by product
Tern’s owner describes community spaces, encrypted messaging, and compact specialist models intended to run locally on phones. Confirm current capabilities and assurance evidence with the product team. Do not infer independent security certification from an architecture description.
Find a relevant starting point
Use our assessment to explore product fit, then discuss scope and availability with the team.
Find your workflow →