Data & your staff
Two questions that get asked as one.
Buyers conflate them, and the answers are different. One is about model training. The other is about what your own company is recording about its staff.
Question one
Is our data training a model?
No. That is Bedrock's guarantee, and we do not change it.
Inference runs in your own AWS environment. Prompts and completions are not used to train the underlying models, and TokenLock adds no destination outside your account — no telemetry, no shared control plane, nothing routed through us. The answer you already got from AWS remains the answer.
Question two
What is our company recording about staff?
Whatever you set retention to, including nothing.
Usage and cost reporting works fully with retention off. Conversation text is only stored if you choose to store it, and it is only read if an administrator runs an extraction.
| Retention | Cost reporting | Extractions |
|---|---|---|
| off | Full | Not available — no text stored |
| 7 days | Full | Last 7 days |
| 30 days | Full | Last 30 days |
| 90 days | Full | Last 90 days |
| 1 year | Full | Last year |
| indefinite | Full | All conversations |
Retention expires the oldest turns first.
If you turn retention on, you are recording your staff's conversations.
That carries legal obligations in most jurisdictions, and they are yours rather than ours: telling people, having a basis for it, keeping it proportionate, and handling access requests. We are not your lawyers. We would rather write this sentence than let you find it out later.
What is recorded per request
Always
person · team · project · application
model · region · profile served by
input, output and cache tokens
cost · timestamp · allowed or refused
Only if you set retention
conversation turns, for the window you chose
read only when an administrator runs
an extraction
There is no keystroke logging, no screen capture, and no per-person productivity score. TokenLock is a spend tool.
Start with retention off.
You get the whole spend picture. Turn it on later, deliberately, once you have decided what you owe your staff.