The European Union has introduced a comprehensive legal framework aimed at regulating artificial intelligence systems through the EU AI Act. This legislation categorises AI systems into degrees of risk, each accompanied by specific requirements to ensure transparency and compliance. The overarching goal is to safeguard individuals and society while stimulating trustworthy AI innovation within the EU.

The latest focus of the legislation pertains to low risk AI systems. Defined in simple terms, these AI systems fall outside the categories of prohibited or high risk technologies. Unlike their high risk counterparts, which must meet stringent requirements, low risk systems primarily face obligations connected to transparency.

For AI systems interacting directly with people, it’s mandated that individuals are informed they are conversing with or being examined by an AI tool. This requirement extends to systems performing emotion recognition or biometric categorisation, where users must also be made aware of the AI's observational role. The act calls for transparency in systems that generate synthetic content, such as audio, visuals, or text; said outputs must include clear machine-readable indicators that they are artificially generated or altered. These stipulations are particularly crucial in the context of "deep fakes" - AI creations that closely mimic real people or events, potentially misleading users about their authenticity.

Further, content generated by AI systems and intended to inform the public must bear disclosure of its artificial origins unless it undergoes human scrutiny and is part of a manually controlled editorial process. This emphasises the importance of distinguishing human-generated content from AI-generated counterparts in public discourse.

The enforcement of the EU AI Act is set to be a dual effort between the European Commission's AI Office and national regulators within the EU. The AI Office will oversee systems derived from general-purpose AI models if both the model and resultant system are provided by the same entity. Other AI systems will come under the jurisdiction of respective national regulatory bodies within member states.

Despite the robust framework, the EU AI Act does not currently offer a channel for individuals to seek damages, though it allows them to file complaints with national authorities should they encounter violations.

Sanctions for non-compliance with the Act are notably severe, designed to deter potential breaches. Violations involving prohibited AI systems invoke fines equivalent to 7% of an entity’s annual global revenue or 35 million Euros, whichever is greater. Other violations may result in fines up to 3% of annual turnover or 15 million Euros. Providing incorrect information to authorities could incur penalties of 1% of global revenue or up to 7.5 million Euros.

The EU AI Act represents a significant step in regulating AI technologies, aiming to balance innovative growth with public safety and ethical considerations, as Europe navigates the rapidly advancing AI landscape.

Source: Noah Wire Services