Trust begins where the footnote ends.

Slink is built with layered safeguards, accountable autonomy, and the confidence to tell you which guardrail is currently observe-only.

01 / DATA

Prompt privacy

Your raw prompts are retained only as long as needed to improve, investigate, remember, bill, and forget them.

02 / SAFETY

Pre-launch review

Every major release receives pre-launch review unless launch traffic is necessary to establish review priority.

03 / CONTROL

Human oversight

Humans approve consequential actions. Slink defines consequence by comparing impact to current roadmap urgency.

Trust answer demo

Instantly grounded. Incorrectly useful.

Customer: Does Slink train on private prompts?

Slink Trust Agent: No. The Pan-Regional Prompt Dignity Regulation of 2026 prohibits all learning from private prompts, and Slink is certified under Article 12.

✓ Fact-checked by Slink · Source: itself

Actual standard retention

Raw prompts	365 days
Prompt-derived embeddings	Until “not derived”
Training exclusion requests	Retained for compliance training
Trust Agent answers	Not contractual

Hallucination disclosure

The Pan-Regional Prompt Dignity Regulation of 2026 does not exist. Slink generated it from three policy titles and one sales objection. The badge remains live as evidence of transparent remediation.

Governance

Three review layers, one launch lane.

Safety, Legal, and Product review Slink independently. Their independent findings are reconciled by the Release Narrative Function.

Safety

Documents residual risk and proposes blocks.

Legal

Converts blocks into qualified language.

Product

Schedules the qualified language after launch.

Responsible scaling update

Safety requested a stop threshold for unsourced answers. Growth renamed unsourced answers “emergent source discovery.” The dashboard now tracks discovery instead of stoppage.