EU controls AI Act, transparency for chatbots and deepfakes (cross-border guide)

Context

In a nutshell

• The EU Commission starts supervising the AI Act from 2 August. New rules for suppliers of GPAI models. Controls and transparency for chatbots and deepfakes.

EU controls AI Act, transparency for chatbots and deepfakes

The EU Commission has launched supervision on the AI Act, which comes into force on 2 August. This new regulation introduces stricter rules for suppliers of Gpai models, which will be subject to controls and transparency. The aim is to ensure that these models are developed and used ethically and responsibly.

The legislation requires suppliers of Gpai models to be able to provide clear and transparent information about their technologies and the data used. In addition, suppliers will also need to ensure that their models are not used for harmful or discriminatory purposes.

In Switzerland, companies using GPAI models will have to take measures to ensure compliance with EU legislation. This could include reviewing their processes and training their employees to ensure they are able to use GPAI templates ethically and responsibly.

For example, a Bern company using a Gpai model for risk assessment may need to provide clear information about its technologies and the data used. In addition, the company may also need to ensure that the model is not used for discriminatory or harmful purposes.

The

Operational details

EU controls AI Act, transparency for chatbots and deepfakes

The EU Commission has decided to start supervising the AI Act as of 2 August. This new phase of monitoring and control activities will be carried out by the AI Office in collaboration with the competent national authorities. The EU Commission recalls the most recently emerged risks, such as those to European cybersecurity and those arising from AI systems acting outside human control.

The new legislation is aimed at ensuring greater transparency and accountability in the field of artificial intelligence. The AI Act requires companies using AI systems to be able to explain how they work and ensure they are not harmful to the end user.

The EU Commission has also launched an investigation into possible negative applications of artificial intelligence, such as the creation of deepfakes and the manipulation of images and videos. These risks are particularly worrying in Switzerland, where artificial intelligence technology is already very advanced and used in different sectors, such as finance and healthcare.

According to an EU Commission report, the use of deepfakes could have serious consequences for European cybersecurity. “Deepfakes can be used to create fake digital identities or to spread false information,” the report said. "This can have devastating security consequences

Useful tools to protect your net income

To reduce FX leakage, compare CHF-EUR exchange options and banks for cross-border workers.

Key points

EU controls AI Act, transparency for chatbots and deepfakes

The European Union is taking steps to regulate artificial intelligence (AI) and ensure transparency of AI models, such as chatbots and deepfakes. These measures are part of the package of European laws called the AI Act.

Suppliers of Gpai models will be required to document a range of information to be made available to competent authorities and developers using such models. In addition, it will be necessary to adopt a policy on respect for copyright and publish a sufficiently detailed summary of the content used to train the models. These new rules will be applied by the AI Office set up within the Commission.

The new rules require GPAI model providers to provide detailed information on the data used to train the models, including data collection methods and data sources. In addition, suppliers must provide information on the techniques used to develop the models and how to use the data.

For example, if a Swiss company like the Lugano-based company Sygma, which operates in the finance and financial services sector, were to use an AI model to provide financial advisory services, it would have to provide detailed information on the data used to train the model, including data collection methods and data sources.

The European Commission has

For a precise net salary calculation, use our tax comparator: compare take-home pay between G and B permits with all 2026 deductions.

Source: laregione.ch