AI safety first
Verifiable AI for regulated work.
AXIX builds vertical AI systems for regulated activities, starting in food safety. We combine practical expertise, traceable sources and professional review.
ResearchWhite paper v6.0
HACCP for AI: a proposal for auditable AI self-control
Where we come from
Our experience comes from food safety work: cold chains, procedures and audits. In this field, a control matters when you can show what was checked, what happened and who was responsible.
We apply that discipline to AI systems: define the intended use, sources, limits and the points where a professional must be able to review the result.
Safety first, in practice
Four commitments that apply to every system we build. They are written to be verified, not just read.
Control points, not good intentions
Every system has its critical points, each with a measurable limit and a corrective action defined before it is needed.
A person before every consequence
No output with consequences takes effect without review by a competent person who has the authority to change it.
Every statement with its source
Answers point to the official text they come from. When there is no source, the system says so.
Data stays with whoever entrusts it
Client content is not used to train models. Each organisation works in a separate environment.
For every product we publish a System Card: what we measure, how, and which limits remain open.
Research · White paper v6.0 · CC BY 4.0
HACCP for AI
In the HACCP for AI white paper, we translate food safety self-control into hazards, control points, limits and corrective actions for AI systems.
The paper proposes three organisational levels and seven control points across the lifecycle. It relates them to ISO/IEC 42001, the EU AI Act and the NIST AI Risk Management Framework.
The seven control points across the lifecycle of an AI system
A
Data acceptance
B
Training-run initiation
C
Capability evaluation
D
Release and scale-up
E
Model and supplier
F
Human review before effect
G
Autonomy and self-modification
Born in food. Serving regulated industries.
Food is the industry we know first-hand. In other regulated fields, we start from their own sources and responsibilities instead of carrying food procedures across unchanged.
Primary industry
Food safety
HACCP plans, hazard analysis, audits against ISO 22000, BRCGS and IFS. The trade our founder has practised since 1997, and the industry of our first product.
Public administration
Official controls, procedures, documentation with sources.
Healthcare and pharma
Processes where traceability and review are mandatory.
Hospitality and contract catering
Hotels, canteens and chains with many sites to keep aligned.
Companies adopting AI
A self-control plan for the AI systems you already use, following HACCP for AI.
Products
HACCP.ai
The first product built with the AXIX method: the workspace for food safety consultants and quality managers. In private beta.
Documents from fixed rules
Plans are built from confirmed facts and verifiable rules. No language model writes them.
A draft until you sign
The professional decides the critical points, reviews and signs.
Company
AXIX® is an Italian company founded by Simone Paciaroni, a food safety consultant and auditor. We design and build in Macerata, Italy.