Auditable AI: what an audit trail of AI reasoning has to show.
Most AI logs record inputs and outputs, not reasons. What GDPR, the EU AI Act and the Product Liability Directive ask for, and why the trace has to be generated, not written afterwards.
Auditable AI is AI whose decisions can be traced: what it decided, from what, and why. Most systems log inputs and outputs only. PRISM, Synapse Net's model, measures its own consistency while it reasons, so every decision comes with a plain-language record generated by the architecture itself.
What most audit trails miss.
Today's AI audit trails record who asked what and what came back. That proves something happened. It does not show why the system decided it, which is the question a regulator, a court or a customer asks.
The explanation is becoming a duty.
GDPR gives people a right to meaningful information about the logic of automated decisions; the EU Court of Justice confirmed in 2025 that this means an explanation a person can understand. From 2 December 2027, the EU AI Act requires high-risk systems to keep automatic logs, be transparent to the businesses using them, and support a right to an explanation of decisions. From 9 December 2026, the new Product Liability Directive treats software and AI as products, with disclosure duties in court.
Why the trace must come from the architecture.
A chatbot can be asked to explain itself, and it will produce a plausible story. Research on reasoning models has shown that the explanation a model writes does not always match what actually drove its answer.
PRISM's record is different in kind: the consistency check is the reasoning step, so the record of what was checked, what contradicted and what was corrected is a by-product of how the answer was reached, not a story written after it.
The same contradiction, the same correction.
When PRISM's energy crosses the threshold, the correction always happens before output. That makes its behaviour on a given contradiction repeatable, which is what an auditor needs.
Short answers
Straight answers to what investors and partners ask us first.
AI whose decisions can be traced and reviewed: what it decided, from what inputs, and why.
For high-risk systems, yes: automatic logging, transparency to deployers and support for explanations, applying from 2 December 2027 for most high-risk uses.
AI that behaves the same way on the same input. For audits, what matters most is that safety behaviour, such as PRISM's correction on a contradiction, is repeatable.
See it working.
We show the PRISM proof of concept live, and share the technical brief and data room with qualified investors and partners.
Nothing on this website is an offer to sell or a solicitation to buy securities.