QuantPi and the EIC Accelerator: Testing AI for Trust

QuantPi, founded in 2020 as a spin-off of the CISPA Helmholtz Center for Information Security and Saarland University and led by managing director Philipp Adamidis, builds a platform that automatically tests AI systems for performance, robustness, bias and other risks and turns the results into evidence for audits and regulations such as the EU AI Act. It was selected in the EIC Accelerator's June 2023 cut-off with a grant-only award of €2.50 million for a €3.57 million project from February 2024 to January 2026 to extend the platform to generative AI, according to CORDIS.

The final report says QuantPi developed its PiCrystal technology to generate test scenarios for generative AI automatically, ran proof-of-value projects with large companies, integrated the platform with "leading cloud providers for continuous monitoring", and reached "the finalisation of our technology as planned", according to the CORDIS reporting page. It says clients could assess generative AI systems ">5-times faster" than with separate assessments and that the platform checks conformance with more than 100 standards and regulations. These are the company's own figures.

Where QuantPi Stands Now

In December 2025 QuantPi joined HPE's Unleash AI partner programme, making its platform available on HPE Private Cloud AI, according to QuantPi, and in March 2026 it joined the NVIDIA Halos AI Systems Inspection Lab ecosystem for testing physical AI such as autonomous vehicles and robots, according to QuantPi. Its customer stories name ControlExpert, which uses computer vision to assess car damage for insurers, according to QuantPi. Its net loss fell to €882,245 in 2025, and a capital increase was registered in June 2026, according to North Data. No new funding round has been announced.

The Technology: Probing the Black Box

Companies deploying AI need to show it is accurate, robust and fair, and under the EU AI Act many will have to document it. Testing each risk separately is slow and hard to apply to large language models. QuantPi's approach perturbs inputs and measures how outputs change, needing only query access to the model, so it can test systems whose inner workings are hidden. The result is a set of metrics and reports aimed at auditors and regulators.

What the EIC Award Promised, and What the Record Shows

The award funded the first automated risk platform for generative AI. QuantPi reports it completed the technology, ran trials with business partners and integrated with cloud monitoring tools. It has since secured distribution through two large technology partners. What it has not shown publicly is revenue, customer numbers or independent validation of its speed claims.

1. Generative AI testing technologyDeliveredFinalised, per report.
2. Industrial proofs of valueDeliveredSeveral partners.
3. Cloud integrationsDeliveredThree, plus HPE.
4. Commercial scaleUnverifiedRevenue not public.

The Money

The EU covered 70% of the project. QuantPi's grant-only award carried no EIC equity. Before the EIC, it received German federal grants of €701,423 in 2021 and €348,733 in 2022, according to North Data, and its register shows capital increases in March and May 2022. Its losses have fallen each year since 2023, but its accounts do not show revenue.

What Cannot Be Checked

Why Europe Should Care

The EU AI Act creates demand for exactly the kind of evidence QuantPi's platform produces. A European company that can test AI systems independently of their makers supports both compliance and trust. QuantPi has built the tool; its partnerships with American hardware groups show where the market is.

The Verdict

QuantPi and the EIC have not been asked for comment for this article. It is based entirely on public records and published statements, linked throughout.