Loading...

28.04.2021

Artificial Intelligence Comparative Guide-Part 2

Ece Özelgin, Special Counsel & Head of Public Policy

Miray Muratoğlu, Associate

6 Employment

6.1 What specific challenges or concerns does the development and uptake of AI present from an employment perspective? How are these being addressed?

One of the main concerns about AI is that it will lead to a loss of jobs due to the deployment of AI applications in the workplace. To address this, Turkey is exploring ways to ensure that its workforce – in particular, younger workers – is trained to use AI technologies in the workplace without this increasing unemployment. Accordingly, the 11th Development Plan (2019–2023) seeks to prepare roadmaps for the development of new technologies, including AI, and the training of the qualified human capital needed for these technologies. In this regard, instead of resisting the adoption of AI, Turkey aims to train its population to better prepare for its use in the employment context. In this regard, the government has introduced its 1 Million Software Developers project, which aims to create employment opportunities by providing free certified online training in several areas, including AI, software/programming, cybersecurity, digital design, personal development, secure internet, telecommunications systems, regulation and orientation.

Meanwhile, the Turkish Labour Code requires employers to comply with the principle of equal treatment in relation to all employees and thus prohibits discrimination based on employees' language, race, sex, disability, political thought, philosophical beliefs, religion and sect or similar grounds. As certain AI-based tools used in recruitment and performance evaluation processes pose a risk to this principle, employers must ensure transparency for these tools. Additionally, as explained in question 4.1, employees and candidates have the right to object to the use of such tools if the results are the sole criterion on which assessments are made, both in recruitment and in performance evaluation.

7 Data manipulation and integrity

7.1 What specific challenges or concerns does the development and uptake of AI present with regard to data manipulation and integrity? How are they being addressed?

AI is designed to conduct calculations that are beyond the capabilities of humans at impressive speeds. It is difficult to explain how AI works, and sometimes even the owners and developers of AI technologies do not understand exactly how they work. This is mainly because AI conducts its calculations in a black box that cannot be accessed by others; only the final product is presented to the public. This poses a threat to data integrity, as we can only trust that the AI is respecting the integrity of the data. If incorrect data is entered on a platform, on the other hand, this breaches the integrity of the data and will skew the results of the algorithm. The AI technology develops by itself through machine learning, meaning that humans cannot intervene in any actions that precede the creation of the invention or command. If there is a breach of data integrity and the AI builds on that breach, the results could be disappointing. This could prove a fatal flaw in sectors such as healthcare, where data about a patient's wellbeing could be wrongly entered, resulting in false diagnosis and treatment. There is no built-in protection against data manipulation. AI can be tricked into reading data incorrectly, which could result in fatal accidents.

8 AI best practice

8.1 There is currently a surfeit of ‘best practice' guidance on AI at the national and international level. As a practical matter, are there one or more particular AI best practice approaches that are widely adopted in your jurisdiction? If so, what are they?

Turkey is in the process of drafting a best practice guide in collaboration with the Scientific and Technological Research Council of Turkey (TUBITAK). The aim is to establish a bridge between the private sector and public policies. The practice areas that are currently under discussion include counter-terrorism efforts, education and the creation of an AI ecosystem. The draft is in its very early stages and was discussed at an Artificial Intelligence and Turkey panel held in 2019. Data protection and privacy, and the training of expert personnel, are the issues at the top of the priority list in relation to the draft. Many AI projects in Turkey are funded by the private sector, underlining the importance of private interest in the forthcoming guidance document.

The transformation of the legal system due to the COVID-19 pandemic is also shaping Turkey's AI landscape, with digital trials now being conducted across the country. This is another area that will be addressed in the guidance document.

The European Commission has also published Ethics Guidelines for Trustworthy AI, which sets out best practice guidelines. These set out seven key principles of trustworthy AI, accompanied by an assessment list. The list is followed by best practice suggestions for each of the seven principles. The guidelines are not necessarily exhaustive, but rather offer a broad idea of best practices and thus cannot be taken as definitive guidance.

8.2 What are the top seven things that well-crafted AI best practices should address in your jurisdiction?

As the head of the Digital Transformation Office of the Presidency of Turkey, Ali Taha Koç, has pointed out, in order to eliminate the ethical and privacy-related concerns associated with AI and to mitigate the adverse effects on society, AI practices should:

  • be human centred;
  • be fair;
  • be transparent;
  • be reliable;
  • be accountable;
  • be value based; and
  • aim to enhance social welfare.

Accordingly, AI practices should take ethical values into consideration – for example, the obligation to create a sustainable, production-based environment.

8.3 As AI becomes ubiquitous, what are your top tips to ensure that AI best practice is practical, manageable, proportionate and followed in the organisation?

While AI practices may vary depending on the context, the use case and the target users, certain technical requirements should always be met in order to guarantee the consistency of the results. In order for AI practices to meet the characteristics outlined in question 8.2, they should ensure robustness and verifiability – that is:

  • the system should not have adverse effects or deviate too much from the baseline behaviour; and
  • the developer should be able to verify the output of the AI.

Companies should prepare their own policies to safeguard accountability and transparency, and embody these principles in their production processes as general principles. As it is easy for developers or stakeholders to forget about the accountability that is associated with the development of AI system, companies should ensure that their employees are regularly trained in these principles and general rules of personal data protection and privacy, as well as competition. Companies should also work on a ‘readable' format that can explain the nature of the systems created to the average person, while also preserving their trade secrets.

9 Other legal issues

9.1 What risks does the use of AI present from a contractual perspective? How can these be mitigated?

As explained in question 1, there is no special regime that governs contractual obligations in relation to autonomous decisions and actions of AI. Thus, in terms of breach of contract, establishing causation in relation to AI products may not be possible if the defect cannot be traced back to human error. This could result in nobody at all being held liable. In addition, if machines gain the ability to negotiate contractual terms and conclude contracts independently, the current legal framework may be insufficient to regulate this.

In order to mitigate these risks, stating them explicitly in the contract could be one option to establish causation.

9.2 What risks does the use of AI present from a liability perspective? How can these be mitigated?

As explained in question 1, there is no special responsibility regime that governs damages caused by the autonomous decisions and actions of AI; such issues must thus be resolved based on general principles of tort law. However, given the special nature of AI, not all of the associated problems can be properly addressed under the current framework, which results in various risks.

The first risk arises from proving fault and causality where damages have resulted from AI. In general, AI systems are not transparent or accountable, which makes it difficult to prove who is actually responsible for the behaviour that has caused damage, and thus both to claim damages and to defend against liability. Similarly, the lack of transparency and accountability may make it difficult to identify the responsible party, especially in cases where multiple parties are involved in producing the AI-based product. Although joint liability may solve this problem to some extent, the allocation of compensated claims among the parties may not be addressed.

To eliminate these risks, the most viable option under Turkish law may be to introduce a new strict liability regime that would hold AI producers and developers strictly responsible for defects and damages that may arise in relation to their AI system, without considering fault or negligence.

9.3 What risks does the use of AI present with regard to potential bias and discrimination? How can these be mitigated?

While the impact of bias will depend on the area in which the AI is used, AI systems are highly controversial in relation to discrimination, as the systems are generally not sufficiently transparent to explain the rationale behind the decision-making process in an understandable manner. However, given that AI applications are not commonly used in the Turkish public sector, concerns are generally limited to practices in the private sector. Accordingly, the most critical issues arise in relation to employment practices, where companies' efforts to automate the recruitment practice may contravene the principle of equal treatment of employees (see question 6.1). If the dataset on which AI is trained is itself biased on the grounds of language, race, sex, disability, political thought, philosophical belief, religion or sect or similar grounds, due to repetitive mistakes and bias already ingrained in society, this will likely result in discrimination against candidates, even though neither the candidate nor the employer may be aware of this.

While individuals have a right to object of the use of such tools pursuant to the Personal Data Protection Law of Turkey, in order to balance the drive towards automation with the mitigation of AI-related risks, AI best practices and ethical standards will play a crucial rule. Efforts to introduce standards and promote the use of tools that increase transparency and accountability should be accelerated and embodied in the National Artificial Intelligence Strategy.

10 Innovation

10.1 How is innovation in the AI space protected in your jurisdiction?

Under the current legislative framework, activity in the AI space is encouraged and incentivised, rather than being restricted, given the limited number of AI-focused innovations that have made it to market. The AI-related policies that have been announced thus far are mentioned in the 11th Development Plan for 2018-2023. Article 346 of the plan states that it will support the development of industrial cloud services for priority sectors. In this regard, technology suppliers will be encouraged to build software and services that can be provided through an industrial cloud platform; and the use of this platform by businesses will be encouraged by funding for its development. Security is addressed in Article 473 of the plan, which states that policies in this regard will be determined by the National Strategy for the Development and Distribution of Technologies for Artificial Intelligence.

10.2 How is innovation in the AI space incentivised in your jurisdiction?

Several incentives are aimed at promoting innovation in the AI space. Accordingly, the Small and Medium Enterprises Development Organization of Turkey's 2019-2023 Strategic Plan aims to promote the dissemination of high technology by enhancing and accelerating the entrepreneurship of small and medium-sized enterprises.

The Ministry of Industry and Technology's 2023 Manufacturing and Technology Strategy recognises that traditional goods and services may soon be replaced by intelligent products and services, along with the spread of disruptive technologies such as IT. To expedite technological advancements in the fields of defence, aviation and space technology, efforts to promote private sector investment and productivity will continue. With regard to innovations such as autonomous vehicles, AI applications and smart weapons systems, strategies and roadmaps will be prepared. Partnerships with foreign companies will be established to facilitate rapid entry into emerging technology markets, such as AI and machine learning, robotics and the Internet of Things. If required, company or technology acquisition opportunities will be evaluated and adequate resources will be allocated.

Another incentive is the Ministry of Justice's Judicial Reform Strategy, which aims to conduct studies on the potential use of AI practices in the judiciary, in line with the standards and guidelines of the Council of Europe, and with the principle of security of legal guarantees.

In the meantime, the Ministry of the Treasury and Finance's New Economy Policy 2020-2022 has introduced initiatives to explore the economic benefits of big data sources. It has been announced that a Big Data and Artificial Intelligence Institute will be established.

11 Talent acquisition

11.1 What is the applicable employment regime in your jurisdiction and what specific implications does this have for AI companies?

Under the Turkish employment regime, the obligations of employers and employees are either:

  • mandated in writing through the individual employment contract, the collective bargaining agreement (if there is a trade union authorised in the related workplace), internal regulations and the workplace policies; or
  • established implicitly by workplace practices and applications.

The labour regulations also have a semi-mandatory nature, whereby:

  • they are non-alterable to the detriment of employees; and
  • individual provisions in employment contracts may be validly applied as long as they are compliant with the labour regulations.

Accordingly:

  • each person employed under an employment contract is entitled to the minimum wage;
  • employees must receive increased payment for overtime work and work for extra hours;
  • in order to employ individuals in Turkey, the employer must be registered with the Social Security Institution (a local presence is required for this purpose);
  • the economic rights (but not the moral rights) to an invention that is created by an employee in the course of his or her employment rests with the employer, with fair consideration due to the employee depending on his or her contribution;
  • employment contracts may be categorised as fixed-term or indefinite contracts;
  • remote working is permitted:
  • contracts may be terminated by either party on the following principal grounds:
    • resignation;
    • mutual termination protocol;
    • termination on justified grounds;
    • termination on valid grounds; or
    • termination without providing valid or justified grounds during the probationary period;
  • an employee who has been employed for at least one year by the same employer becomes entitled to severance pay; and
  • job security is protected and the employee may claim reinstatement in case of unfair dismissal.

11.2 How can AI companies attract specialist talent from overseas where necessary?

To attract talent, companies may avail of the benefits afforded by special zones, which are supported by the government. ‘Free zones' are designated areas in which special regulatory treatment applies to the operation of companies, in order to promote exports of goods and services. Free zones offer a more convenient and flexible business climate in order to increase production and exports for some industrial and commercial activities. To support high-tech products and R&D-based and high-added-value production, specialised free zones have also been established; and in terms of support and incentives, the IT sector is prioritised. The incentives and support available for companies and employees include:

  • different levels of exemptions from taxes including corporate and income tax;
  • exemptions from employees' social security premiums;
  • support at the rate of 50% of the monthly gross wage for up to10 qualified personnel, up to a maximum of $15,000 per employee;
  • support at the rate of 50% of the rental fee in the zones, up to a maximum of $75,000 per year; and
  • interest or dividend support for a maximum of 10 years, as long as this does not exceed 50% of the fixed investment amount they have committed.

A facilitated and special work authorisation procedure is also available for foreign employees, so that they are not subject to the general rules.

12 Trends and predictions

12.1 How would you describe the current AI landscape and prevailing trends in your jurisdiction? Are any new developments anticipated in the next 12 months, including any proposed legislative reforms?

The 11th Development Plan (2019-2023) provides for the conduct of studies at a national scale to promote the production of domestic technologies in the AI field and the dissemination of such technologies throughout the Turkish economy. In this regard, the Digital Transformation Office and the Ministry of Industry and Technology have held a Workshop on National Artificial Intelligence Strategy, focused on transparency, privacy and data security, as well as AI opportunities, the effective use of resources and the transformation of the workforce. While we expect the strategy to be finalised in the next 12 months, the establishment of the Artificial Intelligence Institute – which will implement AI projects and produce information to be considered when setting policies and standards on issues such as the management, protection and dissemination of data – is one initiative envisaged in the strategy.

While no specific legislation on AI is expected, it is anticipated that a framework and criteria for data access and sharing and increasing the efficiency of the use of AI will be introduced. Cybersecurity legislation is also being prepared which may affect the adoption of AI technologies. The legislation is expected to introduce certain data localisation requirements and criteria, so AI models hosted outside Turkey may be restricted due to cybersecurity concerns and the government's efforts to keep Turkish data in Turkey. Lastly, as supporting AI technologies is high on the government's agenda, further incentives and prioritisation may be expected for companies that invest in this area.

Moreover, certain remarks made by Turkish government officials indicate that Turkey will prioritise the security of its national data and its AI algorithms. In this regard, it has also been stated that Turkey's AI strategy must be fair, transparent, reliable, accountable, value-based and dependent on national and ethical values, while also enhancing social welfare.

13 Tips and traps

13.1 What are your top tips for AI companies seeking to enter your jurisdiction and what potential sticking points would you highlight?

The Turkish government is insistent that Turkish data must remain in Turkey. International data transfers are almost impossible to conduct, which could hinder the functions of any AI tools that involve the Internet of Things, a cloud computing system or simple data transactions. Therefore, companies must consider the Personal Data Protection Law thoroughly before embarking on any activity that requires cross-border data transfers.

Turkey further does not recognise AI as the author of an invention in terms of a patent application. This approach in is line with that followed in the United Kingdom and the European Union, where the author must be a human person. The same principle applies with regard to copyright. This is something for AI companies to consider when they are entering the market, as they will need to shape their patent and copyright applications accordingly.

As there is no specific legislation governing AI in Turkey, companies must associate their AI programs with established legal entities before making any applications. Given that as yet there is no legislation in this regard, companies could make use of this opportunity and present highly persuasive arguments regarding their products to the courts.

As explained in question 11.2, many types of support and incentives are available for research and development (R&D) activities conducted in Turkey. In addition to tax exemptions and support for certain types of expenditure, there are programmes aimed at supporting the R&D activities of small and medium-sized enterprises (SMEs) established in Turkey. In this regard, the Small and Medium Industry Development Organization facilitates the commercialisation of R&D products through the following support programmes: