Trust and Security at Aggi

A small company, and plain about it

Aggi Technologies LLC is a small, founder-led company in Plano, Texas, formed in 2008. Your questions go to the people who built Aggi ARIA. Write to prasad@aggicorp.com or call (972) 351-3342.

What we keep, and what we do not

• The key you give us to reach your AI endpoint is encrypted when it is saved, and decrypted only in memory at the moment a test runs. Nobody at Aggi can read it back.
• AI Health Check (Shape B): we do not keep the raw test prompts or your model's raw answers beyond each run. We keep the scores and results, so you can see how behavior changes over time.
• Aggi ARIA: we keep your model's answers to our test prompts, so your compliance team can see exactly what failed and why. Only people you give the compliance officer role, or higher, can view them.
• When Aggi ARIA uses an AI model itself, for example to read documents you upload, we keep only a fingerprint (a hash) of each prompt and answer in the audit trail, never the text.
• We never log your users' live traffic. The one exception is the optional inline mode, and only if you turn on full logging, which is off by default.
• Behavioral audits are read only. Unless you choose the optional inline mode, nothing sits in your request path and nothing changes in production.
• If your data cannot leave your environment, Aggi ARIA can run inside it.
• Card and bank details are handled by Stripe, our payment processor, and never reach Aggi's servers.

Honest limits

• Aggi ARIA gives you evidence about how your AI behaved on the tests we ran. It does not certify that your AI is compliant, safe or free of problems.
• Aggi does not yet hold a SOC 2 report of its own. We say so up front, and we test our own AI use with Aggi ARIA.

Details for your security review

Ask us for where Aggi ARIA is hosted, the outside services it relies on, how long results are kept, how to have your data deleted, and what happens to your data if Aggi stops operating. We answer in writing.