The EC Updates AI Act: Banning Systems That Pose ‘Unacceptable Risk’
The European Commission (EC) has made significant updates to the Artificial Intelligence Act (AI Act), which was originally introduced in 2023 to provide a comprehensive regulatory framework for the development and application of AI technologies. These updates aim to strengthen the safety and rights of individuals within the European Union (EU) by banning AI systems that are categorized as posing an unacceptable risk.
Definition of Risk Levels
The AI Act outlines four distinct levels of risk associated with AI systems:
- Unacceptable risk
- High risk
- Limited risk
- Minimal risk
According to the EC, systems classified as posing an unacceptable risk are now prohibited in the EU. These systems are identified as a clear threat to the safety, rights, and livelihoods of individuals. Examples include:
- Social scoring systems that assess individuals based on their behavior.
- Facial recognition technologies using internet or CCTV material for database creation.
- AI applications that manipulate, deceive, or exploit vulnerabilities.
Monitoring High-Risk AI Systems
While systems identified as high risk are not outright banned, they will be subject to stringent monitoring by the EC. These applications, which are developed in good faith but have the potential to cause significant harm, include:
- AI components in critical infrastructure whose failure could threaten public safety, such as in transportation.
- AI tools in educational settings that may impact access to education, such as exam scoring.
- AI safety components in products, including robot-assisted surgeries.
- AI applications in law enforcement that could infringe on fundamental rights, like evidence evaluation.
Before any high-risk AI system can enter the market, it must undergo a conformity assessment and comply with all stipulations of the AI Act. Additionally, these systems must be registered in a dedicated database and bear the Conformité Européene (CE) Mark, a certification that has governed products within the European Economic Area (EEA) since 1985.
Enforcement and Compliance
The enforcement of the AI Act will be overseen by national regulators. Companies found to be non-compliant with the regulations will face substantial financial penalties, with fines reaching up to €35 million or more, depending on global annual revenue. Furthermore, organizations engaged in prohibited AI practices may be banned from operating within the EU altogether.
Implications for AI Developers
Despite the stringent regulations, most AI developers within the EU may not face these challenges, as the majority of AI systems currently in use are categorized as presenting minimal or no risk. The EC emphasizes that while many AI systems offer valuable solutions to societal challenges, it is crucial to address the risks associated with certain AI technologies to prevent undesirable outcomes.
In summary, the recent updates to the AI Act reflect the EU’s commitment to ensuring that AI technologies are developed and applied in a manner that safeguards the rights and safety of its citizens.
