The EU AI Act Prohibitions Date, 2 February 2025, is the date on which the prohibitions under the European Union's Artificial Intelligence Act (AI Act) took effect. The AI Act, which entered into force on 1 August 2024, establishes a risk-based regulatory framework for AI systems within the EU. The prohibitions target AI applications classified as posing 'unacceptable risk', banning them from the EU market and from use by EU providers or deployers, with limited exemptions. This date is the first of several staggered compliance deadlines under the regulation, which phases in obligations over 6 to 36 months after entry into force.
The AI Act was proposed by the European Commission on 21 April 2021, passed the European Parliament on 13 March 2024, and was unanimously approved by the EU Council on 21 May 2024. The final text was revised to address the rise of generative AI systems, such as large language models and generative AI, whose general-purpose capabilities did not fit the original risk categories. The regulation applies extraterritorially, meaning providers outside the EU must comply if they offer AI services to users within the EU, similar to the General Data Protection Regulation.
Prohibited Practices
The AI Act bans AI systems that are considered to pose unacceptable risks to safety, livelihoods, and fundamental rights. These include:
- Manipulative and exploitative systems: AI that deploys subliminal, manipulative, or deceptive techniques to distort behaviour and impair informed decision-making, causing significant harm.
- Exploitation of vulnerabilities: AI that exploits vulnerabilities of individuals or groups due to age, disability, or socio-economic situation to materially distort behaviour.
- Social scoring: AI used for evaluating or classifying individuals based on social behaviour, personal characteristics, or predicted behaviour, leading to detrimental treatment in contexts unrelated to the original data or disproportionate to the severity of the behaviour.
- Real-time remote biometric identification: AI systems that use real-time biometric identification (e.g., facial recognition) in publicly accessible spaces for law enforcement, except in narrowly defined situations such as searching for missing persons or preventing specific terrorist threats.
- Predictive policing based on profiling: AI that predicts criminal or administrative offences solely based on profiling or personality traits, unless used to support human assessment with objective and verifiable facts.
- Untargeted scraping of facial images: AI that scrapes facial images from the internet or CCTV footage to create or expand facial recognition databases.
- Emotion recognition in certain contexts: AI used to infer emotions in workplaces and educational institutions, except for medical or safety reasons.
- Biometric categorisation for sensitive attributes: AI that categorises individuals based on biometric data to deduce race, political opinions, religious beliefs, or sexual orientation.
These prohibitions are absolute, with no possibility of derogation, except for specific exemptions for law enforcement in the case of real-time biometric identification, subject to prior judicial authorisation and strict time and geographic limits.
Exemptions
Articles 2.3 and 2.6 of the AI Act exempt AI systems used exclusively for military, defence, or national security purposes, as well as systems used for pure scientific research and development. The regulation does not apply to AI systems placed on the market for military purposes or to those used in the context of national security, regardless of the provider's location. Research activities that do not involve placing an AI system on the market are also exempt, but if a research prototype is deployed in a real-world setting, it may fall under the regulation.
Additionally, the Act does not apply to AI systems used for personal, non-professional activities, such as hobby projects or personal assistants, unless they are made publicly available. Open-source AI models are subject to reduced transparency requirements, but the prohibitions on unacceptable risk apply regardless of open-source status.
Timeline and Implementation
The prohibitions took effect on 2 February 2025, exactly six months after the AI Act entered into force. This early deadline reflects the urgency of banning the most harmful uses of AI. Other provisions follow a staggered timeline:
- August 2025: Obligations for general-purpose AI models, including transparency and copyright requirements, become applicable.
- August 2026: Most provisions, including those for high-risk AI systems, become applicable.
- August 2027: Obligations for high-risk AI systems embedded in regulated products (e.g., medical devices, vehicles) become applicable.
Member states are required to designate national competent authorities to enforce the regulation, and the European Artificial Intelligence Board, established by the Act, coordinates national supervision and ensures consistent application across the EU.
Enforcement and Penalties
Non-compliance with the prohibitions can result in fines of up to €35 million or 7% of a company's global annual turnover, whichever is higher. This is the highest penalty tier under the AI Act, reflecting the severity of using banned practices. Smaller providers and start-ups may face reduced fines, but the prohibitions are absolute, and no leniency is granted for ignorance.
The European Commission can conduct investigations and issue corrective measures, including withdrawing AI systems from the market. National authorities can also take action within their jurisdictions. The Act provides for a complaint mechanism, allowing individuals to submit complaints about AI systems to relevant authorities.
Impact on Industry and Innovation
The prohibitions have significant implications for AI developers and deployers, particularly in sectors like AI research, machine learning, and deep learning. Companies must conduct thorough audits to ensure their AI systems do not fall into the banned categories. For example, OpenAI and Anthropic, which develop large language models, must ensure their products are not used for social scoring or manipulative practices. Similarly, Google DeepMind and other research labs must align their work with the regulation.
The ban on real-time biometric identification in public spaces affects companies like Waymo and Tesla Autopilot, which rely on computer vision, though these systems are primarily used in vehicles, not public surveillance. The prohibition on untargeted facial image scraping impacts data collection practices, pushing companies to adopt privacy-preserving techniques.
Despite the strictness, the Act includes exemptions for law enforcement, allowing real-time biometric identification in specific emergency situations, such as terrorist attacks or missing persons cases, with judicial oversight. This balance aims to protect fundamental rights while enabling security measures.
Global Influence and Future Outlook
The EU AI Act has become a global benchmark, influencing AI regulation in other jurisdictions. The prohibitions date sets a precedent for banning AI practices that violate human rights, and other countries may adopt similar measures. As of 2025, the EU is the first major economy to implement such comprehensive AI prohibitions, and its enforcement will be closely watched.
Future developments include the General-Purpose AI Code of Practice, published on 10 July 2025, which provides guidance for providers of general-purpose AI models to demonstrate compliance. The code covers transparency, copyright, and safety, and while participation is voluntary, it is expected to become the de facto standard.
In summary, the EU AI Act Prohibitions Date marks a pivotal moment in AI governance, setting clear boundaries for the use of AI in the EU. It reflects a commitment to human-centric AI, balancing innovation with the protection of fundamental rights. As the regulation is implemented, it will shape the development and deployment of AI technologies worldwide.