Confidence: Watch — Substantive Regulatory Read, But Zero Live European Customers and Limited Release

The EU's decision to delay its high-risk AI rules in June has provided vendors with a significant reprieve. Selling Presence to European banks and insurers would have been nearly impossible after 2 August. Fortunately, that deadline is now off the table. However, another regulation is on the horizon. In just eleven days, the new Article 50 disclosure requirements will come into effect.

Forget the AI Act; that isn't the primary obstacle for this product in Europe and the UK. The real issue lies in how OpenAI operates. They currently lack a single European reference customer, and their data residency plans for Presence remain unclear. Furthermore, their setup allows US-based engineers direct access to core company systems. Procurement teams at banks will halt deals over these three concerns long before regulators get involved.

OpenAI Presence: An Overview

OpenAI Presence was launched on 22 July. It is their new enterprise tool designed for building and managing voice and chat agents. On paper, it offers a wealth of features: built-in guardrails, policy controls, testing tools, and an automated system powered by Codex that attempts to rectify agent errors post-launch.

Setup occurs on a process-by-process basis, such as managing billing or insurance claims. Customers dictate the rules regarding approvals and when to escalate queries to a human. However, you cannot simply visit a website and sign up; it is currently available only in a limited release. OpenAI is deploying its engineers alongside select major tech partners to facilitate the setup for buyers through their sales teams.

OpenAI has named three primary partners: BBVA in Mexico, SoftBank in Japan, and IAG in Australia. However, none of these partners are actively using the tool yet; they are all still in the testing phase. Notably, none are based in Europe.

---

So What

The Regulatory Cliff Everyone Expected Has Moved

Had this launched a month and a half ago, it would have encountered a significant regulatory barrier on 2 August. Fortunately, that threat has dissipated for the time being. EU lawmakers have postponed the major high-risk AI deadlines. Standalone systems now have until December 2027 to comply, while AI integrated into existing regulated products has until August 2028. Parliament and the Council approved the final text last month, and the delay officially took effect in July.

This delay offers OpenAI a considerable advantage, even if they did not earn it. Credit scoring is included on that high-risk list. If a bank employs an agent, like BBVA's, to assist in credit decisions, it qualifies as a high-risk system. The same applies to technology that determines insurance prices or assesses risk. The only real difference now is that vendors have an additional sixteen months before these rules become enforceable.

Want to read more?

Subscribe to sowhatnowwhat.co.uk to keep reading this exclusive post.

Subscribe Now