Powerful AI needs human governance.
American AI's Trust architecture combines user protection, privacy, independent assurance, enforceable boundaries, incident response, evidence and a human-authored moral floor.
Personalisation is not the same thing as conscience.
A useful personal AI should adapt to the person it serves. It can learn preferred style, priorities and ways of working. It can also exercise judgment — reasoning, warning, challenging and disagreeing when needed.
But American AI separates those choices from protected, human-authored fundamental principles that are not simply removable through ordinary user settings.
Different controls for different problems.
These are cooperating layers. They are not alternative names for one feature.
Helps the user assess suspicious material, risky choices and recovery steps.
Governs information, memory, permissions and exposure to providers or other systems.
Independent scrutiny of evidence, authority, contradictions and control integrity.
Turns safety decisions into enforceable system boundaries.
Coordinates incident response, recovery and continuity when something has materially gone wrong.
Matches effort and resources to consequence while preserving required quality and safety.
Defines deeper human-authored principles, supervisory authority and responsibility.
AUDI is deliberately separate from ordinary operational work.
AUDI is the current customer-facing identity of the independent PIA AUDITOR function. It belongs in the governance and assurance IP story rather than as another consumer product. Its role remains evidence review, authority checking, contradiction detection and receipt verification.
The separation matters: the system doing the work should not be the only system deciding whether that work was authorised, internally consistent and properly evidenced.
The threat also comes from outside the assistant.
Generative AI is making fraud, impersonation and social engineering more convincing and scalable. PIA Protect expresses the defensive side of the thesis: people increasingly need intelligent assistance on their side.
The proposition is assistance and resistance — not a promise of perfect detection or immunity.
People need safe AI on their side.
Evidence, approvals, receipts and recovery make action inspectable.
Trust is stronger when consequential work leaves a clear trail: what was requested, which authority applied, what was approved, what executed, what evidence was retained and how recovery occurs if something goes wrong.