Le shadow IAor "shadow AI" or "clandestine AI", is a specific form of shadow IT applied to the field of artificial intelligence.
Shadow IA refers to the use and deployment of artificial intelligence technologies within an organisation without the knowledge, approval or supervision of IT departments or formal management bodies.
Why does shadow AI exist?
Several factors are contributing to the emergence of Shadow AI:
- Easy access to AI tools : The market today is full of AI platforms in SaaS (Software as a Service) mode, open source code libraries and AI APIs (Application Programming Interfaces), which are often highly intuitive and accessible even to non-IT specialists. It has become extremely simple for a business user to sign up to an online AI service, use an chatbot or integrate predictive analysis functionality into a spreadsheet, without the need for in-depth technical skills or IT endorsement.
- Need for agility and rapid innovation: In an increasingly competitive business environment, business teams are under constant pressure to innovate quickly, improve efficiency and meet the specific needs of their departments. AI is seen as a powerful lever for achieving these objectives. Faced with IT approval processes that are sometimes considered long or complex, teams may be tempted to bypass formal channels to deploy AI solutions more quickly and directly, in order to gain greater autonomy and responsiveness.
- Ignorance or lack of understanding on the part of IT: Sometimes, the IT department itself is not yet fully familiar with AI technologies, or has not yet put in place clear policies and procedures to govern their use. This can create a vacuum, encouraging business teams to take the lead and experiment with AI on their own, without necessarily feeling supervised or held back by IT.
- Pressure to solve specific business problems: Business teams are often faced with very specific and urgent challenges that may seem secondary or less of a priority to the IT department, which has a more global view of the company's challenges. AI can appear to be the ideal solution for addressing these specific problems quickly, and teams may be tempted to implement AI solutions 'on the sly' to demonstrate their value and achieve concrete results quickly.
The advantages and disadvantages of Shadow AI
Shadow AI is not intrinsically good or bad. It has potential benefits, but also significant risks.
✔ Benefits
- Innovation and increased agility : Shadow AI can encourage experimentation, rapid innovation and the discovery of new AI applications within the company, by enabling business teams to test solutions without bureaucratic obstacles.
- Rapid response to specific business needs: It allows you to solve specific and urgent business problems more quickly and flexibly, by adapting AI to the specific needs of each department.
- Discovering talent and innovative ideas: Shadow AI can reveal talented and creative employees who have innovative ideas for using AI, and who might not have had the opportunity to express themselves in a more formal setting.
❌ Disadvantages and major risks
- Security and compliance risks : the use of AI tools that are not approved and supervised by IT can lead to security breaches, breaches of confidentiality and breaches of confidentiality. sensitive dataand regulatory non-compliance (RGPDetc.). The data used by shadow AI may not be properly protected, the tools used may be vulnerable, and the processes put in place may not comply with the company's security and compliance standards.
- Lack of control and governance : the lack of centralised supervision makes it difficult to control the use of AI, manage risks and coordinate initiatives. This can lead to a proliferation of non-coherent, non-integrated and potentially redundant or contradictory AI solutions.
- Integration and compatibility problems : shadow AI solutions, developed in isolation, can be difficult to integrate with a company's existing information systems, which can limit their effectiveness and create data and information silos.
- Unnecessary expenditure and wasted resources: lack of coordination can lead to duplication, unnecessary purchases of software licences, and wasted financial and human resources.
- Ethical risks and bias algorithmic : Without proper supervision, Shadow AI solutions can unintentionally incorporate algorithmic biases, lead to unfair or discriminatory decisions, and raise important ethical issues.
💡 Solutions for businesses and IT departments
Shadow IA poses a major challenge to businesses and IT departments. It highlights the need to adapt and modernise IT governance approaches, and to strike a balance between innovation and control.
For IT departments, it is crucial to :
- Become aware of the existence of shadow AI : recognise that Shadow IA is a growing reality and not ignore it.
- Understanding the motivations of shadow AI : Analyse why business teams use Shadow IA, and what their needs, frustrations and expectations of IT are.
- Establish clear and agile AI governance: define clear policies and procedures to govern the use of AI, while encouraging innovation and experimentation. The aim is not to ban shadow AI, but to channel it and make it safer and more controlled.
- Propose alternatives and solutions: offer business teams approved and secure AI platforms, tools and services that meet their needs and are easy to use.
- Educating and raising awareness : train employees in the risks and good practices associated with the use of AI, and make them aware of the importance of AI governance.
- Working with business teams: Establish an open and constructive dialogue with business teams, to understand their needs, support them in their AI projects, and together build an approach to AI that is responsible and beneficial to the business.
📊 Figures and statistics
In France
- Informal use of AI tools
According to an IFOP-Talan study (2023) quoted by several media, around 16 % of French people use Generative AIAnd 68 % of them do not inform their superiors. An article in LeMagIT reports that 68 % of French employees use generative AI applications (such as ChatGPT) outside the framework authorised by their company, a clear illustration of the 'Shadow GPT' phenomenon. - Adoption by company size
According to a BPI France-Le Lab study (March 2024):- 28 % of large SMEs (100-249 FTEs) use these tools,
- 20 % of SMEs,
- 15 % of VSEs
🌍 Worldwide
- Use not authorised in the United States
An Ipsos study for Reuters revealed that in the United States, 28 % of employees surveyed regularly use ChatGPT, while only 22 % have official authorisation for this type of use. - Global adoption and a growing market
While the Shadow AI phenomenon is manifesting itself in various countries, the global generative AI market is growing rapidly (estimated at $454 billion in 2023 with annual growth of around 19 %). This context is encouraging many employees to adopt these tools in an unsupervised way to gain in productivity, even if this entails risks (data security, compliance with the RGPD, etc.).
