Curvestone is named an OpenAI Select Partner

The OpenAI Partner Network is a global programme for organisations that build, sell and deliver AI solutions with OpenAI. It brings together partners with industry expertise, delivery capability and existing customer relationships, and gives them resources, enablement and support so that enterprises can adopt OpenAI frontier models and products and turn them into measurable results.

Select Partner is a tier within that network. For Curvestone it means continuing to work with OpenAI to help organisations build, deploy and scale AI responsibly, with earlier access to model capability and the support to put it into production properly.

We are a compliance company, not a model company. We do not train frontier models and we have no intention of starting. Our work is the layer above: taking the best available models and making them do regulated work that stands up to supervision. Being closer to the provider of those models directly improves the product our customers use.

"Being named an OpenAI Select Partner recognises the work we are doing to put the best available AI models to work in one of the most heavily regulated corners of financial services. Compliance is where AI has to be right, not just fast, and partnering with OpenAI lets us give mortgage, lending and advice firms the confidence to automate regulatory work without cutting corners," said Dawid Kotur, Founder and CEO of Curvestone.

Most AI use cases tolerate a wrong answer. A summary that misses a nuance costs a reader thirty seconds. A case file review that misses an affordability inconsistency costs a firm a redress exercise and an accountable individual a difficult conversation with the regulator.

That asymmetry changes the engineering. It means every finding has to carry its basis: what was read, what rule was applied, what evidence supports the conclusion. It means staged model release with automated evaluation before anything reaches a live customer. It means an audit trail on every case rather than a confidence score. None of that comes free with a model. It is the work built around one.

Curvestone automates the compliance checks regulated firms already run, for mortgage brokers, networks, lenders and wealth advisers operating across the UK. The live use cases are the recognisable ones.

• Full case file reviews. The whole evidential bundle read and cross-referenced against the firm compliance checklist: fact finds, income evidence, bank statements, identification and suitability letters.
• Financial promotions. Checking adviser and firm marketing against the financial promotion rules before it goes out, rather than after a complaint arrives.
• Financial crime controls. Checks applied consistently across cases instead of unevenly across reviewers.

The deployment model matters as much as the checking. We have written about why the check has to run inside the case journey rather than in a separate portal a broker has to remember to open.

The result firms care about is coverage. A proper file review takes hours, adviser numbers keep growing, and compliance headcount does not scale with them. So most firms sample. They review a fraction of new business and accept the risk on the rest, because the arithmetic of doing anything else does not work.

A platform that can review every case rather than a sample moves a firm from spot-checking a fraction of files towards full coverage. That de-risks the business and it changes what compliance is for. Instead of a cost centre that slows growth, it becomes the function that lets a firm scale volume without adding headcount or risk, with an audit-ready evidence trail on every case.

We have made that argument at length elsewhere, including on why the Mills Review turns compliance into an advantage and on why AI gives compliance teams time back rather than replacing them.

Three things. We will deepen how we use OpenAI models, including the reasoning and cost characteristics of the current generation, so that each check does more useful work per token. We will broaden the range of checks the platform can run. And we will bring the automation to more firms across the sector.

Alongside that, the governance keeps pace. Curvestone is certified to ISO 27001:2022 and to ISO/IEC 42001:2023, the international standard for AI management systems, both under certificate number 19087 issued by ISOQAR. Our data stays in the UK and EEA on Microsoft Azure with zero data retention on model calls. The certificates, the AI Policy overview and the latest CREST accredited penetration test summary are all requestable from the Curvestone Trust Centre.

Better models are only useful in this sector if the governance around them is as good as the model. That is the part we are responsible for.