← Services

Virtova services · By Sultan Meghji

EU AI Act readiness

EU AI Act readiness for U.S. banks, insurers, healthcare firms, and PE portcos with European exposure: classification, technical documentation, governance, and the audit trail Brussels actually expects.

A lot of U.S. boards have been told the EU AI Act is a European problem. For most internationally active U.S. banks, and for a meaningful share of mid-size U.S. firms with any cross-border footprint, that framing is wrong in a specific and expensive way.

The Act, Regulation (EU) 2024/1689, applies extraterritorially. A non-EU firm whose AI system produces output used in the EU is within scope for the relevant obligations, regardless of where the firm is incorporated and regardless of whether the firm has European staff. Most U.S. financial institutions with European customers, operations, or partners meet at least one of those triggers through at least one product line. The high-risk classifications around credit scoring, employment, and essential-services access map directly onto common U.S. banking, insurance, and healthcare use cases.

The sharper operational point is that the Act forces a documentation discipline with no clean U.S. analogue yet. Auditors moving through a U.S. firm’s EU-facing workflows are already asking for that documentation, and supervisors increasingly expect a written EU AI Act posture as a baseline. Most U.S. firms underestimate this until they meet it. Building the documentation set once and applying it both ways is the engagement Virtova exists to scope.

Engagements are led personally by Sultan Meghji. Sultan’s tenure as inaugural Chief Innovation Officer of the U.S. FDIC included direct engagement with European supervisors (the European Central Bank, the German Bundesbank, the United Kingdom Ministry of Defence, and EU institutions) on the cyber and AI agenda; the work has continued in advisory and boardroom settings since.

What this engagement looks like

A Virtova EU AI Act readiness engagement typically runs eight to twelve weeks. The standard scope covers five threads.

Inventory and classification. Each in-scope AI system is classified against the Act’s categories: prohibited, high-risk (Annex III), limited-risk, minimal-risk, or general-purpose AI. Classification is the rate-determining step; everything else flows from it. We expect, and clients sometimes resist, the finding that several systems they did not consider AI are in fact in scope.

Article-by-article gap analysis. For high-risk systems, the obligations span risk management (Art. 9), data and data governance (Art. 10), technical documentation (Art. 11), record-keeping (Art. 12), transparency (Art. 13), human oversight (Art. 14), accuracy/robustness/cybersecurity (Art. 15), and quality management (Art. 17). Output: a written gap list, prioritized by enforcement risk and business criticality.

Documentation build. Technical documentation is where most U.S. firms have the largest gap. We build the set the Act actually expects, not a generic compliance memo, to a standard that an external auditor or competent authority will accept.

Governance integration. EU AI Act compliance is wired into the firm’s existing risk and AI governance program rather than built as a separate workstream. For firms with U.S. exposure, the same engagement also produces NIST AI RMF mapping so a single program serves both rulebooks.

Phase-in tracking. The Act phases in across 2024–2027. Different obligations land at different times. We build a tracker the firm carries past the engagement so deadlines do not arrive as surprises.

When the engagement is the wrong answer

EU AI Act readiness is the wrong scope when the organization has no plausible EU exposure: purely domestic U.S. firms with no European customers, partners, or operations and no foreseeable plan to acquire any. It is also the wrong scope when the gap is foundational AI governance rather than EU-specific obligations; in that case AI governance consulting is the right starting engagement, with EU readiness as a follow-on.

Where the Act overlaps with sector-specific EU regulation (DORA for financial services, MDR for medical devices, GDPR throughout), Virtova engagements scope the AI Act layer specifically and bring in named specialist counsel for the adjacent regimes when the engagement warrants it.

A note on enforcement

Enforcement of the Act is a national competent-authority responsibility, with EU-level coordination through the AI Office. Penalties are material: up to €35 million or 7% of worldwide annual turnover for the most serious violations. The first wave of enforcement decisions in 2026 has set expectations that competent authorities are taking documentation discipline seriously. Firms that have built the documentation report short conversations. Firms that have not report less short ones.

The pattern across early enforcement is also worth flagging for U.S. risk committees. The questions competent authorities are asking are largely documentation questions, not technical ones. Where the technical artifact is in place, the conversation moves quickly. Where it is not, and the firm cannot produce a coherent risk-management or technical-documentation file at the article level, the conversation becomes a discovery exercise that a U.S. firm running on a different documentation culture is not built to handle.

Next step

Most engagements start with a 30-minute discovery call. Bring an honest read of the firm’s EU exposure (customers, operations, partners, output-flow) and we will tell you which classification work needs to happen first and what a focused engagement looks like.

"Most U.S. boards have been told the EU AI Act is a European problem. For most internationally active U.S. banks, this is wrong in a specific and expensive way."
— Sultan Meghji

Frequently asked

What is EU AI Act readiness?
EU AI Act readiness is the work of getting an organization's AI systems, governance, and documentation into a state that satisfies the obligations of Regulation (EU) 2024/1689, the European Union's risk-based AI regulation. The Act applies extraterritorially: a non-EU firm whose AI output is used in the EU is in scope for the relevant obligations.
Does the EU AI Act apply to U.S. banks?
Often yes. The Act applies to any provider or deployer of AI systems where the output is used in the EU, regardless of where the firm is incorporated. Most internationally active U.S. banks meet this trigger through at least one product line. High-risk classifications around credit scoring, employment decisions, and essential-services access map directly onto common U.S. banking, insurance, and healthcare use cases.
What are the key obligations under the EU AI Act?
The Act takes a risk-based approach. Prohibited AI practices are banned outright. High-risk systems carry obligations around risk management, data governance, technical documentation, transparency, human oversight, accuracy, robustness, and cybersecurity. Limited-risk systems carry transparency obligations. General-purpose AI models carry their own documentation and copyright-disclosure obligations. The phase-in runs from 2024 through 2027.
What does a Virtova EU AI Act readiness engagement produce?
A classification of each in-scope AI system against Act categories, a written gap analysis against the relevant high-risk obligations (risk management, data governance, technical documentation, human oversight, accuracy and robustness, cybersecurity), a remediation roadmap with named owners, and a documentation set that will satisfy an external auditor or competent authority. Where Virtova does not have native EU-counsel depth, named outside counsel is brought in by reference and disclosed.
How does EU AI Act readiness interact with NIST AI RMF compliance?
The two frameworks overlap meaningfully but are not identical. NIST AI RMF is voluntary and process-oriented; the EU AI Act is binding and prescriptive about specific obligations and documentation. Firms with both U.S. operations and EU exposure typically build a single underlying governance program that satisfies the more prescriptive Act obligations and maps cleanly to NIST functions. Virtova's engagements are scoped to do that.

Related Virtova services

Related writing

Work with Virtova

Most engagements start with a 30-minute call.

Confidential by default. NDAs available on request.

Book a discovery call →