EC Outlines AI Transparency Goals

The European Commission has set the obligations companies will face when new AI transparency rules come into play next month. The purpose of those rules is to tackle threats of misrepresentation.

It is the next stage in a long-running and controversial process which ultimately saw the EC concede some ground on areas involving biometrics and critical infrastructure. The EC published guidelines explaining that AI companies are obliged to let people know when they are dealing with the technology and use machine-readable marks to spot content made or manipulated by it.

There are also requirements to let people know when they are shown deepfakes along with AI-generated content involving matters of public interest where no human oversight is involved. The same warnings are to cover emotion recognition or biometric categorisation systems. It is also offering guidance on how to apply the requirements, for example by explaining direct interaction covers services including chatbots.

The EC is also supplying information on how companies can prove they are meeting the transparency requirements of the AI Act, pointing to a code of practice that provides legal certainty. A majority of the AI Act is scheduled to be implemented on 2 August, including the enforcement powers of the EC and relevant national authorities. The EC reiterated that systems already available have until 2 December to comply with the requirements around marking and detecting content.