Accountability is the product thesis
Enterprise AI has a proof problem. Models are deployed without published track records, decisions are influenced without owners, and outcomes go unscored. Prophesee exists to close that gap, so accountability is not a compliance section on our website; it is the product thesis. The commitments below follow from it.
The EU AI Act and Prophesee
The EU AI Act's transparency obligations are in force, and its high-risk obligations apply from December 2027 (standalone systems) and August 2028 (embedded systems). Prophesee is designed so that the evidence those obligations demand accumulates as a by-product of operating the platform, rather than being assembled retrospectively.
- Documentation and record-keeping. Every model carries a Model Passport: its version history, validation record, backtests with misses included, approvals, and scored outcomes. The record is generated by operation, with a date on every entry.
- Human oversight. Decisions carry named owners. Machine outputs route through thresholds and exceptions to people with authority to act, and every override is recorded with its reason.
- Accuracy and robustness. Predictions are backtested on frozen data before deployment and scored continuously against outcomes in production. Degradation is a detected event, not a retrospective discovery.
- Logging and traceability. Every prediction, decision and outcome is bound on one event backbone, so "which model influenced this decision, on what evidence, approved by whom" is a lookup, not an investigation.
- Deployment control. Prophesee runs on your premises, on your data, inside your permission model.
We track the Act's implementation timeline and evolving standards, and we build toward them ahead of their dates. Where our customers' systems fall into high-risk categories, the platform's records are theirs to use as evidence.
We write with AI
We write with AI. The essays, films and pages we publish are drafted with AI tools, including Anthropic's Claude. What makes the work ours is everything around the drafting: the arguments are ours, every statistic is verified against a named source by a person, every piece is edited and approved by a person, and editorial responsibility sits with 3RDi.
When the subject is Prophesee itself, the drafting is grounded in the platform's own internal documentation, so what we publish describes the platform as it stands, not a roadmap.
AI-generated text increasingly carries provenance signals such as watermarks. We leave them intact.
AI in how we operate
The same standards we sell are applied to ourselves. The models we use in our own operations, including those inside our compliance and screening workflows, carry the same passports, validation records and human-override paths we provide to customers. Regulators increasingly evaluate how companies govern their own AI; our answer is the same artefact we give our customers.
Ask for the evidence
If your assessment, procurement or compliance process needs more than this page states, ask us. Written answers, with evidence attached. Contact us.