Legislation for chatbots?

New legislation is being drafted for the use of data for the development, training and deployment of AI applications, such as chatbots. In this article we look at the most important trends: where are the laws and regulations going?

1. What is it

Companies that use AI applications – for example to support marketing, sales or customer service processes – must take into account the arrival of additional laws and regulations. This regulation is necessary to prevent abuse, for example, but it also has consequences for the quality of the solutions. In this tech update explanation from a lawyer and a scientist.

In TechUpdate, Ziptone's section on relevant emerging technology, we briefly explain what it is, how it works, why it's relevant, what the pitfalls are and ask the reader's opinion: Hot or Not?

2. Why is it relevant

Companies invest a lot of time and money in developing successful applications of artificial intelligence (hereinafter: AI). The European Commission recognizes that this can contribute to the competitiveness of companies. That is why the European Commission has been opting for a number of years to promote and strengthen European cooperation in the field of AI. However, in the view of the European Commission, citizens' confidence in AI must also be improved and safeguarded.

All kinds of applications of AI can be found in customer contact. Think of speech recognition, which can be used for the automated recording of spoken summaries of conversations; open-ended speech recognition for routing to the right employee or monitoring sentiment; offering real-time next best action and all kinds of other forms of agent assist. Or consider the automated acceptance (or rejection) of new customers based on personal data entered online. And of course, customer data is continuously used to train (improve) AI applications.

Algorithms must not be discriminatory and must respect existing privacy rules. According to the Commission, the existing legislation and regulations are not sufficient to properly manage the application of AI. That is why the European Commission is developing additional regulations. The Dutch cabinet agrees: “We must ensure that the rules that apply in the physical world also apply online. We embrace the opportunities that digitization offers us, but on the basis of public values that are important to us.” said Alexandra van Huffelen, State Secretary for Kingdom Relations and Digitization.

3. How does it work

The new regulations from the EU relate to both hardware (such as self-driving cars) and software solutions (such as chatbots). The first parts of the guidelines are published last spring. The regulations are based on risk assessments. For example, the European Commission has classified AI around chatbot solutions as applications with a relatively low risk. But other online tools may carry a higher risk. Think of an algorithm that determines whether or not a customer will receive a loan or insurance. Incidentally, the current GDPR already imposes many restrictions on what companies are allowed to do with data and algorithms: profiling, in which the algorithm decides without human intervention, but with legal consequences, is emphatically not allowed.

Also DNB, De Nederlandsche Bank endorses the European vision on AI. “For example, a bank chatbot can help you with all sorts of things, such as what to do if you have lost your debit card. A chatbot is also helpful when taking out a loan. This is possible because more and better data are available to assess risks. In addition, a robot with AI can make connections in large amounts of information and recognize complicated patterns. It is important that AI does not discriminate or that the process can no longer be controlled. That is why DNB is working with other supervisors and the sector on guidelines to counteract this.”

In addition to new laws and regulations, there will probably also be a reporting obligation for serious incidents and defects involving AI, including a system of fines, comparable to the fine system for data leaks, where a percentage of the global annual turnover can be imposed as a fine.

4. Pay attention

In all likelihood, a distinction will be made between data use for improving (service) processes via AI and data use for training AI applications. AI applications developed by the business itself – such as low code chatbots – must also comply with current laws and regulations. Companies will have to expand their governance so that low code development work is properly monitored by the business.