Artificial intelligence has spread through enterprise tools faster than the frameworks meant to govern it. Automated features, models embedded in SaaS platforms, augmented processing: AI now operates within organizations’ daily workflows, often without its contours being legible to procurement departments, compliance teams or business decision-makers.
The question facing CIOs is therefore no longer whether the company uses AI. It has become more precise: which systems are involved, with what data, for what purposes, at what level of risk and with what oversight guarantees. This is exactly the question an AI Disclosure answers. This document is shifting from voluntary best practice to contractual prerequisite, under the dual pressure of regulation and procurement departments.
What is an AI Disclosure?
An AI Disclosure is a structured statement through which an organization makes visible how it uses and governs AI. It covers the number of AI systems deployed, the governance frameworks applied and a concise policy statement. Any organization that develops, purchases or deploys AI systems should produce one.
The value of the exercise lies in its granularity. A general statement of intent serves little purpose in a procurement or audit process. A useful AI Disclosure goes down to the level of facts: which systems are embedded in a platform, how their risk level is qualified, which data flows are involved, which indicators track their performance. It is this precision that turns a surface-level commitment into something a third party can actually work with.
The difference with a simple ethical commitment
Many companies already publish an AI charter or ethical principles. These texts have their uses, but they remain at the level of discourse. An AI Disclosure stands apart because it is verifiable: it lists real systems, their classification, their human oversight mechanisms and a commitment to updates over time. It does not describe an intention, it documents a state of affairs.
This logic mirrors that of non-financial reporting. A company is not asked to claim it is responsible; it is asked to produce information that allows its stakeholders to judge for themselves.
Why the pressure is mounting now
Two converging movements explain why the topic is landing on CIOs’ desks this year rather than another.
The EU AI Act entered into force on August 1, 2024, establishing the world’s first horizontal regulatory architecture for AI, built on a four-tier risk classification. Its application is staggered. Prohibited practices have applied since February 2025, obligations for general-purpose AI models since August 2025.
One point deserves attention because it has evolved recently. On May 7, 2026, the European Council and Parliament reached a provisional political agreement, as part of the Digital Omnibus, postponing the application of obligations for high-risk systems: standalone systems under Annex III are now expected by December 2, 2027, and AI embedded in regulated products under Annex I by August 2, 2028. By contrast, the transparency obligations under Article 50, notably for chatbots, remain largely on their original timeline, with a deadline of August 2, 2026. The postponement therefore concerns high risk, not transparency.
The co-legislators’ message is consistent on one point: this additional time is a preparation window, not an exemption. The most demanding part of compliance is not drafting standard documentation; it is the inventory of all AI systems present in the organization and their classification, a task that does not get easier with time.
Added to this is the environmental dimension. The CSRD requires large companies to disclose their environmental impacts, digital and AI workloads included. Measuring and then documenting the footprint of AI usage is thus becoming a reporting topic in its own right.
The market is not waiting for the law
The second movement comes from buyers themselves. Transparency on AI usage is taking hold as a condition of trust between organizations, regardless of the regulatory deadline. The G7 Digital Ministerial meeting reinforced this orientation by reaffirming the need for AI that is safe, trustworthy and mindful of its environmental footprint.
In practice, a procurement department evaluating a SaaS solution will increasingly ask its vendor to document its AI usage. The AI Disclosure becomes a due diligence instrument: a way to objectify practices, verify a platform’s alignment with security and compliance requirements, and narrow the gap between the commercial promise and operational proof.
This is why the title of this article does not describe a distant hypothesis. If your company sells services to large accounts, the demand will come through the contractual channel before it comes through the regulatory one.
The parallel with privacy regulation, and its limits
The comparison with major privacy frameworks is instructive. Like PIPEDA in Canada or the CCPA in California, the EU AI Act has extraterritorial reach: it applies to any organization whose AI systems are used by or affect people located in the Union, regardless of where the provider is established. Like those frameworks in their time, it turns a best practice into an obligation and makes documentation a defensible asset in the event of an audit.
The parallel has its limits, however. Privacy regulation dealt with a relatively circumscribed object, personal data. AI governance touches a broader and more fluid perimeter, where systems evolve quickly and where Shadow AI, those tools used without validation from IT or compliance teams, complicates the inventory. An AI Disclosure is therefore not managed as a one-off compliance exercise, but as a continuous process: document, publish, improve, enrich.

What a CIO can do right now
Without waiting for the 2027 and 2028 deadlines, several workstreams are already actionable and will determine the quality of any future disclosure.
The first is the inventory. Identifying the AI systems actually deployed, including those that took hold through usage rather than formal decision, is the foundation of everything else. The second is classification: qualifying each system by its risk level and purpose makes it possible to distinguish controlled areas from zones of uncertainty. The third is establishing human oversight and monitoring indicators, which give the disclosure its evidentiary value. The fourth is integrating the environmental footprint of AI usage into the existing reporting framework.
These steps are not merely defensive. An organization that masters its AI portfolio makes better infrastructure decisions, cuts unnecessary spending and reports credibly to its investors and clients.
How Fruggr approaches the topic
Fruggr, the AI governance and digital sustainability platform published by Digital4Better, has released its own AI Disclosure to give concrete form to this requirement. The approach is openly framed as evolving rather than definitive: a foundation meant to progress with usage, standards and regulation, rather than a fixed document presented as beyond reproach.
Beyond its own disclosure, Fruggr provides the open observatory AI Pulse, which makes it possible to track AI regulations by jurisdiction, consult an AI register template and submit one’s own disclosure to join a public repository. For an IT department, it is an entry point for situating its own maturity before structuring a complete approach.
The challenge, for CIOs, is not to produce the perfect document on the first attempt. It is to start documenting, because transparency is built over time and the market, for its part, has already started asking the question.
Discover Fruggr’s AI Disclosure or submit your own directly at https://ai-pulse.fruggr.io/en