The EU has started enforcing major parts of its Artificial Intelligence Act from August 2, giving regulators broader powers to oversee advanced AI systems and penalize companies that fail to follow the new rules.
The AI Act was adopted by the EU in 2024 and is being introduced in different stages. It follows a risk-based approach, meaning AI systems that present higher risks to public safety, health, or fundamental rights face stricter legal requirements than lower-risk technologies.
One of the biggest changes affects AI-powered chatbots. Companies must now clearly inform users whenever they are interacting with an AI system instead of a human. The rule is designed to improve transparency and help people understand when artificial intelligence is being used.
The law also requires companies to make AI-generated or AI-edited content easier to identify. Deepfakes and certain AI-generated material related to matters of public interest must include clear labels. However, artistic, fictional, satirical, and creative works are exempt from some of these disclosure requirements.
AI systems that were already on the market before August 2 have until December 2, 2026, to meet the new technical standards for marking and detection. The rules will not apply retroactively, meaning AI-generated content created before the deadline does not need to be labelled.
The AI Act also bans several practices considered harmful. These include AI systems designed to manipulate people, predict an individual’s criminal risk, commonly known as predictive policing, and emotion recognition systems used in workplaces or educational institutions. Most of these restrictions have already been in force since February 2025.
The EU has also expanded the powers of the European Commission’s AI Office. The office can now supervise general-purpose AI models, including advanced systems that power popular chatbots and other large AI applications.
The AI Office is made up of legal, technical, and economic experts. It can request documents, inspect AI models, require corrective actions, and ask companies to provide access to their technology. If serious risks remain unaddressed, regulators can restrict the use of an AI model across the European Union.
The stronger enforcement powers come after earlier challenges faced by regulators while seeking access to Anthropic’s Mythos AI model. National authorities will continue supervising smaller AI systems under the new legal framework.
Companies that violate the AI Act could face significant financial penalties. Businesses involved in prohibited AI practices may receive fines of up to €35 million or 7 percent of their annual global revenue, whichever amount is higher.
Other violations, including providing false information or refusing to cooperate with investigators, can lead to fines of up to €15 million or 3 percent of worldwide annual revenue.
More AI rules are expected later this year. From December 2, the EU will ban AI systems that create non-consensual sexually explicit deepfakes, including so-called nudification tools. The decision follows growing international concerns over AI-generated explicit images.
The EU has also delayed some compliance deadlines for high-risk AI systems. Stand-alone AI used in areas such as biometrics, education, employment, migration, and critical infrastructure must comply by December 2, 2027. AI integrated into regulated products, including toys and lifts, will have until August 2, 2028, to meet the new requirements.
In other related news also read EU Orders Meta to Remove Addictive Facebook, Instagram Features
The phased rollout of the AI Act reflects the EU’s effort to balance innovation with public safety while creating a clearer regulatory framework for artificial intelligence across member states.
