They are 'intended to foster trust and integrity in the information ecosystem.'
Rawf8/Getty Images
The European Union has announced that it's new AI transparency rules are now in effect and enforceable across the bloc as of August 2nd. Article 50 of the AI Act requires content publishers to mark and label certain types of AI-generated content, as well as informing users when they are interacting a AI chatbot or agent rather than a real person.
"These obligations are intended to foster trust and integrity in the information ecosystem," the European Commission wrote in its new guidelines. "People should know when they are interacting with AI or exposed to AI-generated content. This will help them make informed decisions, calibrate their trust and reliance on AI and avoid mis information or deception."
The new rules require an "AI" mark if machines assisted in the creation of "authentic-looking" deepfake content (image, audio, video) or published text. If the image, audio, video or text is generated fully by AI, it must carry an "AI-Generated" label. For the latter, regulators cite examples like "fully AI-generated deep fake videos with politicians or fictional events," "full AI-composed music or art" or "AI-generated news summaries."
Other types of AI must also be labeled, like models that recognize your emotions or process biometric data. "The goal is simple: whenever AI plays a vital role, you are entitled to know," the EC said in a video explainer.
The Commission will oversee providers of general-purpose AI models as well. Those organizations will be required to document information, publish training data summaries and implement copyright policies, with non-compliance subject to enforcement by national authorities.
The new law has some teeth, with fines up to €15 million or three percent of global annual revenue for companies and €750k for EU institutions, bodies and agencies. Proportionality will be taken into account for small and medium-sized companies and small mid-cap companies. The rules exempt personal content like group chats along with "evidently artistic" satirical and fictional works.
The new rules make Europe the most prominent regulator of AI technology. They have drawn criticism from the industry, which has argued that the rules are an undue burden which will slow innovation.
However, Europe's tech regulator stated that the rules will simplify development for AI providers by creating a single market with harmonized rules. "It also addresses potential risks to people's health, safety, and fundamental rights, while safeguarding democracy and the rule of law," the EC wrote. "Ensuring its effective implementation is now a key priority for the Commission."
