A Comparative Perspective Between Europe and the United States
Abstract
The increasing adoption of generative artificial intelligence has transformed the relationship between organizations and software, creating new governance challenges that extend beyond traditional regulatory compliance. While existing frameworks such as the European Union’s AI Act and the NIST AI Risk Management Framework establish important legal and organizational principles, they provide limited guidance on how AI systems should behave during everyday professional workflows.
This article explores the concept of AI Directives as a potential organizational governance layer designed to bridge the gap between regulatory requirements and operational practice. Rather than presenting AI Directives as an established standard, the article examines them as a conceptual framework for documenting behavioural principles, promoting methodological consistency and supporting human oversight across professional environments.
The discussion compares current regulatory approaches in Europe and the United States, analyses the distinction between software governance and behavioural governance, and considers how organizations may progressively develop living operational methodologies that evolve through practical experience. Finally, the article reflects on the possible future standardisation of behavioural governance as artificial intelligence becomes increasingly integrated into professional decision-making.
The objective is not to propose a definitive governance model, but to contribute to the broader discussion on how organizations can translate regulatory principles into consistent, accountable and professionally aligned AI practices.
Contents
1. Introduction
The rapid adoption of generative artificial intelligence has fundamentally changed the relationship between humans and software. Unlike traditional applications, which execute predefined instructions written in programming languages, generative AI systems produce dynamic outputs that depend not only on their underlying models but also on context, user interaction and operational guidance.
This evolution has created a new governance challenge. While governments and regulatory bodies have concentrated on the legal responsibilities associated with artificial intelligence, organizations increasingly face a more practical question: how should an AI system behave during a professional workflow?
The European Union has responded through the Artificial Intelligence Act, introducing a comprehensive legal framework based on risk classification, transparency obligations and human oversight. In the United States, the approach has developed along a different path, relying on voluntary governance frameworks such as the NIST AI Risk Management Framework, sector-specific regulation and a growing body of state legislation. Although these approaches differ in structure, both pursue a common objective: ensuring that AI systems are deployed responsibly.
Yet an important operational question remains largely unanswered.
Regulatory frameworks establish obligations for providers, deployers and organizations. They define accountability, documentation requirements, risk management processes and, in certain cases, mandatory human oversight. However, they generally do not prescribe how a generative AI system should conduct itself while assisting a lawyer drafting an appeal, an engineer preparing technical documentation or a physician organizing clinical information.
This distinction is significant. Traditional software derives its behaviour primarily from source code. Generative AI, by contrast, continuously interprets natural-language instructions throughout its interaction with users. Its operational behaviour is therefore influenced not only by the underlying model but also by the guidance that governs each stage of the dialogue.
As organizations increasingly integrate generative AI into regulated professional environments, a new layer of governance begins to emerge. This layer does not replace legislation, nor does it modify the underlying model. Instead, it seeks to establish structured behavioural rules that define how AI should interact with professionals, when human validation should occur, what methodological principles should be followed and how organizational standards should be preserved across different AI platforms.
This article explores that emerging concept. It examines the current regulatory landscape in Europe and the United States, identifies the operational space that existing governance frameworks only partially address, and considers whether structured AI Directives may become an important component of organizational AI governance during the coming years.
Rather than proposing a specific implementation or product, the discussion focuses on a broader question: whether professional organizations will eventually require standardized behavioural directives to complement existing legal and technical governance frameworks.
2. Traditional Software vs. Generative AI
For decades, software governance has been built around a relatively stable assumption: software behaves according to the logic embedded in its source code. Although defects, security vulnerabilities and unexpected interactions may occur, the intended behaviour of a traditional application is ultimately determined by the instructions written by its developers.
This paradigm has shaped the legal and technical frameworks governing software development. Source code can be reviewed, tested, version-controlled and audited. Changes are introduced through software updates, and every new release represents a deliberate modification of the system’s behaviour.
Generative artificial intelligence challenges this traditional model.
Unlike conventional software, a generative AI system does not simply execute predefined logical branches. Instead, it produces responses by interpreting natural-language instructions within a given context. Two users interacting with the same underlying model may obtain substantially different outputs depending on the prompts they provide, the conversation history, the operational instructions supplied by the organization and the constraints imposed during the interaction.
The model itself may remain unchanged while its observable behaviour varies considerably.
This distinction has important implications for governance.
In traditional software, behavioural consistency is achieved primarily through programming. In generative AI, behavioural consistency increasingly depends on operational guidance. The same model may act as a legal assistant, a medical documentation tool or a customer support representative without any modification to its internal architecture. What changes is not the model itself, but the framework within which it operates.
This does not mean that source code has become irrelevant. The underlying model, its architecture and its technical safeguards remain fundamental. However, they no longer represent the sole mechanism through which behaviour is controlled.
Generative AI introduces an additional operational layer where behaviour is continuously influenced by natural-language instructions, organizational policies and human interaction.
This dynamic nature distinguishes AI systems from traditional applications in a fundamental way. Behaviour is no longer entirely defined before execution begins. Instead, it is progressively shaped throughout the interaction itself.
As a result, governance can no longer focus exclusively on software development. It must also address how organizations define acceptable behaviour during deployment.
This shift may prove to be one of the defining characteristics of the generative AI era. While previous generations of software were primarily governed through programming, modern AI systems increasingly require governance through operational principles that remain effective regardless of the underlying model or technology provider.
3. The Current Regulatory Landscape
The rapid expansion of artificial intelligence has prompted governments and regulatory bodies to develop new governance frameworks aimed at balancing innovation with safety, accountability and public trust. Although jurisdictions have adopted different regulatory philosophies, a common trend has emerged: AI is increasingly treated not merely as a technological innovation, but as a technology capable of influencing fundamental rights, economic activity and professional decision-making.
3.1 The European Union: A Risk-Based Regulatory Framework
The European Union has adopted one of the world’s most comprehensive approaches through the Artificial Intelligence Act (AI Act). Rather than regulating artificial intelligence as a single category, the AI Act classifies AI systems according to the level of risk they present.
This risk-based methodology distinguishes between prohibited AI practices, high-risk systems, transparency obligations applicable to certain AI applications and general-purpose AI models. The obligations imposed on organizations vary depending on the intended use, the potential impact of the system and the role performed by each actor within the AI value chain.
The European model places particular emphasis on documentation, traceability, risk management, transparency and human oversight. Organizations deploying AI are increasingly expected not only to comply with technical requirements but also to demonstrate that appropriate governance mechanisms exist throughout the lifecycle of the system.
The AI Act therefore reflects a broader regulatory philosophy: artificial intelligence should be subject to governance structures that are proportionate to the risks generated by its deployment.
3.2 The United States: A Distributed Governance Model
The United States has followed a markedly different path.
Rather than adopting a single comprehensive federal AI statute comparable to the AI Act, the American regulatory landscape consists of a combination of federal guidance, sector-specific regulation, state legislation and voluntary standards.
One of the most influential initiatives is the NIST Artificial Intelligence Risk Management Framework (AI RMF). Instead of creating legally binding obligations, the framework provides organizations with practical guidance for identifying, assessing, managing and governing AI-related risks. Its emphasis lies in organizational processes, accountability and continuous improvement rather than prescriptive legal requirements.
Alongside the NIST framework, various federal agencies have issued AI-related guidance within their respective areas of competence, while several states have introduced legislation addressing issues such as automated decision-making, consumer protection, employment practices and algorithmic accountability.
This decentralized approach reflects the broader characteristics of the American regulatory system, where governance often develops through a combination of legislation, administrative guidance, industry standards and evolving judicial interpretation.
3.3 Different Models, Common Objectives
Despite their structural differences, the European and American approaches share several fundamental objectives.
Both recognise that artificial intelligence requires more than technical performance. Organizations are expected to understand how AI systems operate, identify foreseeable risks, maintain appropriate documentation and ensure meaningful human oversight where necessary.
Both also acknowledge that governance is not a one-time exercise. AI systems evolve, organizational practices change and new risks emerge as technology becomes integrated into everyday professional activities.
The principal distinction lies not in the objective, but in the regulatory method. Europe relies primarily on harmonised legislation, while the United States places greater emphasis on governance frameworks, sectoral regulation and organizational responsibility.
Yet neither approach fully addresses one practical question that organizations increasingly face on a daily basis:
How should a generative AI system behave while assisting professionals in carrying out their work?
4. The Governance Gap
Regulatory frameworks define responsibilities.
They establish who may be considered a provider or a deployer, identify categories of risk, require technical documentation and, in certain circumstances, impose obligations concerning transparency, record keeping and human oversight.
These requirements are essential. Without them, organizations would lack a coherent legal framework for the responsible deployment of artificial intelligence.
However, governance does not end with regulatory compliance.
As generative AI becomes integrated into everyday professional activities, organizations face operational questions that extend beyond the scope of legislation.
A law firm may require that AI identifies missing facts before drafting legal arguments.
A hospital may require AI to separate clinical observations from diagnostic hypotheses.
A financial institution may require AI to distinguish verified information from assumptions before producing an internal report.
An engineering firm may require every technical recommendation to be supported by recognised standards.
These behavioural expectations are rarely defined by legislation.
Nor are they typically embedded within the underlying AI model itself.
Instead, they arise from the professional methodology, internal policies and quality standards of each organization.
This distinction is significant.
Regulation establishes what organizations are expected to achieve.
Organizations must determine how AI should behave in order to achieve those objectives consistently.
This operational layer becomes particularly important because generative AI does not simply execute predefined instructions. It continuously interprets natural-language input while producing responses in real time.
Consequently, governance is no longer limited to controlling software before deployment. It increasingly involves influencing behaviour during deployment.
This represents a subtle but important shift.
Traditional software governance focused primarily on controlling the application.
Generative AI governance increasingly requires organizations to influence the interaction.
The challenge is therefore no longer limited to technical reliability or legal compliance. It also concerns methodological consistency.
Two professionals within the same organization may use the same AI model and receive technically accurate responses while producing work that differs substantially in structure, reasoning, terminology or analytical depth.
From a regulatory perspective, both interactions may satisfy legal requirements.
From an organizational perspective, however, they may fail to satisfy internal quality expectations.
This illustrates a distinction that has received comparatively little attention within current AI governance discussions.
Legal compliance and operational consistency are not necessarily the same objective.
An organization may fully comply with applicable regulation while still lacking a coherent methodology governing the day-to-day use of generative AI.
Conversely, a carefully structured operational methodology cannot replace legal compliance, but it may strengthen governance by promoting consistency, accountability and professional standards across different users, departments and AI platforms.
As AI adoption continues to expand, organizations are likely to devote increasing attention not only to regulatory obligations but also to the practical question of how AI should participate in professional decision-making.
This emerging operational dimension may ultimately become one of the defining characteristics of mature AI governance.
5. What Is an AI Directive?
If organizations increasingly need to define how generative AI should behave during professional workflows, an obvious question follows:
Where should those behavioural expectations be documented?
One possible answer is through what this article refers to as an AI Directive.
For the purposes of this discussion, an AI Directive may be understood as:
A structured operational document that defines the behavioural principles, methodological rules and interaction standards governing the use of generative AI within a specific organizational context.
Unlike legislation, an AI Directive does not establish legal obligations.
Unlike the underlying AI model, it does not determine how language is generated.
Instead, it operates at the organizational level, translating professional expectations into structured operational guidance.
An AI Directive may address questions such as:
- How should AI distinguish between verified facts and assumptions?
- When should the system request additional information before producing an answer?
- Which sources should be prioritised?
- What level of certainty should be expressed?
- When should human validation be explicitly recommended?
- Which professional style or methodology should be followed?
These questions are not unique to any particular profession.
A law firm, a hospital, an engineering consultancy and a public administration may all answer them differently while pursuing the same objective: ensuring that AI supports professional work in a manner consistent with the organization’s standards.
An AI Directive therefore does not attempt to replace professional judgement. Nor does it seek to automate responsibility.
Its primary purpose is to promote methodological consistency.
Rather than requiring every individual user to develop personal prompting techniques, the organization defines a shared operational framework that reflects its own procedures, quality standards and professional values.
This distinction is important.
The directive governs the interaction, not the professional.
Human users remain responsible for their decisions, interpretations and final outputs. The directive simply establishes the behavioural framework within which AI assistance is expected to operate.
In this sense, an AI Directive occupies a position between regulation and technology.
Regulation establishes the legal environment.
The AI model provides the technical capability.
The directive defines how that capability should be applied within a particular professional context.
Because it exists independently of any specific AI provider, the same directive could, in principle, be adapted for different language models or future AI systems without altering its underlying methodological principles.
This portability may become increasingly valuable as organizations seek to preserve consistent governance while technological platforms continue to evolve.
Whether AI Directives ultimately emerge as an accepted governance practice remains uncertain. However, they offer a useful conceptual framework for understanding how organizations might bridge the gap between legal compliance and day-to-day operational behaviour in the age of generative AI.
6. Why AI Directives Matter
The value of an AI Directive does not lie in making generative AI more intelligent. Modern language models are already capable of producing sophisticated analyses, drafting complex documents and responding to a wide range of professional questions.
The challenge is different.
Organizations do not simply expect AI to produce plausible answers. They expect it to produce answers that are consistent with their own professional methodology.
This distinction becomes increasingly important as AI moves from occasional experimentation to routine use within professional environments.
Consider a law firm where several lawyers use the same language model to prepare legal documents.
Without shared operational guidance, each professional may interact with the system differently. One lawyer may request extensive legal authorities before reaching a conclusion, while another may prioritise brevity. One may distinguish carefully between established facts and assumptions, while another may not explicitly separate them. A third may ask the AI to challenge the proposed legal strategy before producing a draft.
None of these approaches is necessarily incorrect.
However, they may produce work that varies significantly in structure, methodology and analytical consistency despite relying on the same underlying AI model.
The same situation can arise in many other sectors.
A hospital may require AI to organise patient information according to established clinical protocols before generating summaries.
An engineering consultancy may require every recommendation to identify the applicable technical standards and clearly distinguish between mandatory requirements and professional judgement.
A financial institution may require AI to identify uncertainty explicitly whenever available information is incomplete.
A public administration may require AI to maintain a neutral tone while distinguishing legal provisions from internal administrative practice.
These examples illustrate a common principle.
Organizations often seek consistency not because regulation requires identical outputs, but because consistency supports quality assurance, internal training, accountability and professional confidence.
An AI Directive provides a structured mechanism for expressing those expectations.
It transforms individual prompting habits into organizational methodology.
This transition has practical advantages beyond consistency alone.
New employees can begin working within an established operational framework without first developing extensive prompting experience.
Teams can collaborate more effectively because AI interactions follow common methodological principles.
Quality assurance processes become easier to design when behavioural expectations are documented rather than assumed.
Periodic reviews can focus on improving a shared framework instead of correcting isolated prompting practices.
Perhaps most importantly, organizational knowledge becomes less dependent on individual users.
Experience accumulated through repeated interaction with AI can be incorporated into the directive itself, allowing improvements to benefit the entire organization rather than remaining with the professionals who originally developed them.
In this way, an AI Directive functions not merely as a set of instructions for an artificial intelligence system, but as a repository of organizational experience.
It captures methodological decisions that have proven effective over time and makes them available to future users in a structured and reproducible manner.
For organizations operating in regulated or knowledge-intensive environments, this may prove to be one of the most significant advantages of behavioural governance: the ability to transform individual experience into institutional methodology without altering either the underlying AI model or the applicable legal framework.
7. AI Directives as Living Documents
Unlike traditional procedural manuals, AI Directives should not be regarded as static documents.
They operate within an environment that evolves continuously. Artificial intelligence models improve, regulatory expectations change, professional methodologies mature and organizations accumulate experience through daily interaction with AI systems.
For these reasons, behavioural governance cannot reasonably be expected to remain unchanged over time.
An AI Directive should instead be understood as a living document.
Its initial version is unlikely to represent a definitive methodology. Rather, it establishes a structured starting point from which the organization can observe how AI performs in practice, identify recurring issues and progressively refine its behavioural framework.
This process resembles continuous quality improvement more than traditional software development.
Organizations do not merely discover technical defects. They also identify methodological opportunities.
A legal team may conclude that AI should request additional factual information before proposing legal arguments.
An engineering consultancy may decide that technical standards should always be cited before recommendations are made.
A medical department may determine that differential diagnoses should never be presented without clearly identifying the clinical evidence supporting each possibility.
These refinements do not arise because the AI model has changed.
They arise because the organization has learned how AI can be integrated more effectively into its professional practice.
Experience therefore becomes a governance asset.
Every recurring ambiguity, every successful interaction and every methodological improvement provides information that can be incorporated into subsequent versions of the directive.
Over time, the document evolves from a collection of operational instructions into a structured expression of the organization’s accumulated professional judgement.
This evolutionary process also highlights an important distinction.
The objective is not to adapt the organization to the behaviour of AI.
Rather, it is to adapt the behavioural framework governing AI to the evolving needs of the organization.
In this sense, AI Directives should be viewed as subject to continuous review rather than periodic replacement.
Small refinements may prove more valuable than infrequent comprehensive revisions.
As professional experience grows, behavioural guidance becomes progressively more precise, more predictable and more closely aligned with organizational methodology.
Artificial intelligence itself may also contribute to this refinement process.
By analysing previous interactions, identifying inconsistencies or suggesting alternative formulations, AI can assist professionals in reviewing existing directives.
However, the role of AI remains advisory.
The decision to modify a directive—and the responsibility for its content—continues to rest with the organization and the professionals responsible for its governance.
This distinction preserves an essential principle of responsible AI governance.
Artificial intelligence may help improve the behavioural framework under which it operates, but it should not become the authority that defines that framework.
The governance relationship therefore remains clear.
Humans establish the methodology.
AI assists in its continuous improvement.
The revised methodology subsequently governs future AI interactions.
Viewed from this perspective, an AI Directive is more than an operational document.
It becomes an institutional learning mechanism.
Rather than allowing experience to remain fragmented across individual users, the organization captures that experience, evaluates it, and incorporates it into a shared methodological framework that evolves alongside both professional practice and technological progress.
8. Human Oversight Revisited
Human oversight has become one of the central principles of contemporary AI governance.
Both regulatory frameworks and professional guidance frequently emphasize that artificial intelligence should remain subject to meaningful human supervision, particularly where its outputs may influence important decisions.
The principle is widely accepted.
Its practical implementation, however, is less straightforward.
Traditionally, human oversight has often been understood as reviewing the output generated by an AI system before it is used or communicated. In many professional settings this remains both necessary and appropriate. Lawyers review legal submissions before filing them. Physicians validate clinical information before making decisions. Engineers approve technical documentation before implementation.
Output review is therefore an essential safeguard.
Yet it may not represent the only form of meaningful oversight.
As generative AI becomes embedded within everyday workflows, organizations may increasingly exercise supervision at an earlier stage—by defining the behavioural framework under which AI operates.
Instead of asking only whether a particular response is acceptable, organizations may also ask whether the AI was instructed to approach the task in an appropriate manner from the outset.
This distinction reflects two complementary forms of governance.
The first focuses on supervising individual outputs.
The second focuses on supervising the methodology that shapes those outputs.
Neither approach replaces the other.
Even a carefully designed behavioural framework cannot eliminate the need for professional judgement in individual cases. Likewise, reviewing every response without establishing common methodological principles may lead to inconsistent practices across an organization.
Meaningful human oversight may therefore operate at multiple levels.
Professionals remain responsible for evaluating AI-generated work before relying upon it.
At the same time, organizations may assume responsibility for defining the operational principles that guide AI interactions across teams, departments and professional activities.
Viewed in this way, oversight becomes more than a reactive control mechanism.
It also becomes a proactive governance activity.
Organizations do not merely detect errors after they occur. They actively shape the conditions under which AI is expected to operate, reducing unnecessary variability while preserving professional discretion where it is genuinely required.
This layered understanding of human oversight may become increasingly relevant as AI systems evolve.
Future governance is unlikely to depend solely on reviewing individual responses. It may also require continuous attention to the behavioural frameworks, organizational methodologies and professional standards that influence those responses long before they are generated.
Ultimately, effective oversight is not achieved by choosing between human judgement and artificial intelligence.
It is achieved by ensuring that each performs the role for which it is best suited.
Artificial intelligence assists with analysis, organisation and language generation.
Human professionals remain responsible for defining objectives, exercising judgement and accepting accountability for the decisions that follow.
9. Towards Standardisation?
Most organizational practices follow a familiar pattern.
They often begin as informal methods developed by individual professionals. As experience accumulates, organizations document those methods, establish internal procedures and eventually adopt common standards that facilitate consistency, quality assurance and interoperability.
Artificial intelligence governance may follow a similar trajectory.
Today, organizations are experimenting with a wide variety of approaches to guiding the use of generative AI. Some rely on internal policies. Others develop prompt libraries, workflow documentation or practical guidance for employees. These initiatives differ considerably in structure and terminology, reflecting the early stage of organizational AI adoption.
Whether these practices will gradually converge into more formal behavioural frameworks remains an open question.
If they do, standardisation is unlikely to require identical directives across every organization.
Professional methodology is inherently context-dependent. The behavioural expectations of a hospital differ from those of a law firm, an engineering consultancy or a public administration. Standardisation, therefore, would not necessarily imply uniformity of content.
Instead, it may involve the emergence of common structural principles.
Organizations could eventually adopt shared approaches to questions such as:
- how behavioural guidance should be documented;
- how directives should be reviewed and approved;
- how version control should be maintained;
- how modifications should be recorded;
- how human responsibility should be allocated;
- and how organizational learning should be incorporated into successive revisions.
These structural principles would resemble governance frameworks rather than technical specifications.
They would not prescribe professional judgement.
They would define how organizations manage that judgement when integrating generative AI into their operational processes.
Professional associations, industry bodies and standards organizations may also contribute to this evolution.
Just as many sectors have developed model policies for information security, quality management and data protection, future guidance could include recommended structures for documenting behavioural governance. Such documents would not replace organizational autonomy, but they could promote greater consistency, facilitate implementation and encourage the sharing of good practices across professions.
International standards may likewise influence this development. As AI governance frameworks continue to mature, organizations are likely to seek practical mechanisms that connect regulatory obligations with day-to-day operational practice. Behavioural directives, or comparable governance instruments, could represent one possible response to that need.
At present, however, this remains a developing field.
There is no universally accepted model for documenting behavioural governance in generative AI, nor is there a consensus regarding the terminology that should be adopted. The concepts discussed in this article should therefore be understood as a contribution to an ongoing discussion rather than as a description of established international practice.
The future of AI governance will almost certainly extend beyond legislation and technical safeguards alone.
It will also depend on how organizations translate regulatory principles into consistent professional behaviour.
Whether this is ultimately achieved through AI Directives or through other forms of behavioural governance remains to be seen. What appears increasingly clear is that organizations will need mechanisms capable of connecting legal obligations, professional methodology and everyday interaction with generative AI systems.
10. Conclusion
Generative artificial intelligence has introduced a significant shift in the relationship between organizations and software.
For decades, software governance focused primarily on controlling code, managing technical risks and ensuring that applications performed as intended. While these objectives remain essential, generative AI introduces an additional dimension. The behaviour observed by users is no longer determined exclusively by software architecture. It is also influenced by the operational guidance provided during interaction.
Current regulatory frameworks have made substantial progress in addressing this new reality.
The European Union’s AI Act establishes a comprehensive legal framework based on risk management, transparency and accountability. In the United States, governance has developed through a combination of voluntary frameworks, sector-specific regulation and organizational responsibility. Together, these approaches provide an increasingly robust foundation for the responsible deployment of artificial intelligence.
Yet governance within organizations extends beyond legal compliance.
As AI becomes integrated into everyday professional activities, organizations must increasingly determine not only whether AI may be used, but also how it should assist professionals in carrying out their work.
This article has suggested that behavioural governance deserves greater attention within that discussion.
The concept of AI Directives has been presented not as an established international standard, but as a possible organizational mechanism for documenting methodological expectations, promoting operational consistency and capturing institutional experience. Whether this particular terminology ultimately gains acceptance is less important than the broader principle it represents: the need to define behavioural frameworks that complement existing legal and technical governance.
Perhaps the most significant characteristic of such frameworks is that they are not static.
Professional experience evolves. Technology evolves. Regulation evolves.
Consequently, behavioural governance must also evolve.
Organizations may increasingly find value in treating operational guidance as a living body of knowledge—one that is continuously reviewed, refined and strengthened through practical experience while remaining subject to human judgement and organizational responsibility.
The history of technology demonstrates that governance rarely develops in a single step.
Programming standards, cybersecurity frameworks, quality management systems and data protection practices all emerged gradually as organizations sought practical methods for managing increasingly complex technologies.
Artificial intelligence is unlikely to be different.
The coming years may therefore be shaped not only by advances in AI models, but also by advances in the organizational methodologies that govern their responsible use.
Ultimately, the long-term success of artificial intelligence in professional environments may depend less on whether systems become more capable, and more on whether organizations become better at defining how those capabilities should be applied.
The future of AI governance may therefore be determined not only by better models, but also by better methodologies.
Be the first to comment